THE POTENTIAL IMPACT OF BROADBAND COMMUNICATION NETWORK TECHNOLOGY ON CONSUMER MARKETING COMMUNICATION: A COMPUTER SIMULATION EXPERIMENT Dissertation for the Degree of Ph. D. MICHIGAN STATE UNIVERSITY MARTIN PAUL BLOCK I975 I I III III III: II II II IIII II II I m R A I I Michigan State Urfivcrsit'y This is to certify that the thesis entitled THE POTENTIAL IMPACT OF BROADBAND COMMUNICATION NETWORK TECHNOLOGY ON CONSUMER MARKETING COMMUNICATION: A COMPUTER SIMULATION EXPERIMENT presented by Martin Paul Block has been accepted towards fulfillment of the requirements for Ph.D. Mass Media degree in mm Major professor Date October 17, 1975 0-7 639 8.9g! BMCI-r :I/ NC‘ III ' I LIBRARY BI NOERS III III srnlusront mcmu' II ABSTRACT THE POTENTIAL IMPACT OF BROADBAND COMMUNICATION NETWORK TECHNOLOGY ON CONSUMER MARKETING COMMUNICATION: A COMPUTER SIMULATION EXPERIMENT By Martin Paul Block The basic purpose is to assess the potential impact of broadband communication network (BCN) technology on consumer marketing communication using a computer simulation model of human time allocations.IfBCN technology represents the combination of ingeractive cable television, computer and information processing technology, and terrestial communication satellites) The computer simulation model is used in an experimental mode to determine the impact of a hypothetical BCN system on the allocation of time to daily human activity. (Marketing communication is described in terms of a number of traditional marketing activities, including personal selling, advertising, packaging, point-of- purchase, direct mail, product sampling, publicity and public relatigné) A typology of consumer marketing communication is then developed using a diagrammatic model of communication with a source, message, channel, Martin Paul Block receiver, and feedback. (The first type of marketing communication is advertising and promotion which involvesfl¥¥ the passive consumption of messages by large numbers of consumer receivers through the mass mediaP> The second type of marketing communication is retailing and selling which involves the active participation of both the seller and consumer and is generally a face-to-face situation. The third type of marketing communication is marketing research which can be described as the feedback mechanism from the consumer to the marketer. he impact of BCN technology on advertising would center on the addition of response capability and a larger program offering which should result in more attractive program- ming commanding a larger share of the viewer's time. The impact of BCN technology on retailing and selling would include both a reduction in shopping-related travel time due to the substitution of communication for transpor- tation and an increase in the time devoted to overall shopping activity because of the reduction in the cost of shopping. The impact of BCN technology on marketing research is not sufficiently tangible to warrant con- sideration in terms of the allocation of time to daily activity .5 The means of describing the impact of BCN technology is found in the methodology of technological forecasting borrowed from military planning. Martin Paul Block Technological forecasting can be divided into four methodological groups, dialectical methods which include the scenario approach and the Delphi Method, teleological methods including PERT analysis or normative backcasting, experimental and empirical methods including time series analysis, and analytical methods including the method used here in the computer simulation experiment. The computer simulation model developed here, named TIMMOD, allows for the vicarious experimentation with various aspects of human time allocation to daily activities, including the hypothesized impacts already hypothesized as a result of the installation of a BCN system. The empirical basis for TIMMOD is the United States Time Use Survey conducted in 1965. TIMMOD constructs time allocations to various daily activity categories from the selection of random deviates from two basic theoretical probability distri- butions. The first probability distribution describes the activity selection behavior of the hypothetical population, and the second probability distribution describes the duration of the selected activity. The simulation begins with a single individual, who has a randomly selected waking time and retiring time. According to the appropriate hour of the day, an activity is selected, followed by the computation of the appropriate activity duration. The process Martin Paul Block continues until the entire day is exhausted. When the first individual has retired, then the process starts again with a new individual. The activity selection distribution consists of a matrix of cumulative per- centages by activity category and hour of the day. Uniform random numbers are drawn and compared to this matrix to select a particular activity. The activity duration is determined from a family of Erlang distri- butions with different parameters to describe different shaped distributions for different activity categories. The TIMMOD experiments demonstrated impact on the human allocation of time to daily activity categories in that the amount of time allocated to particular activity changed in response to an experimental manipu- lation of various treatment variables. Examining the experimental treatments each in isolation provides some interesting results. Reducing the amount of time spent in shopping-related travel, for example, reduces the amount of time spent shopping and increases the amount of time spent in home and family activities. Increasing the amount of time spent viewing television reduces the amount of time spent participating in other mass media consumption and reduces the amount of overall leisure time. Increasing the amount of shopping time, that is making shopping a less expensive activity, reduces semi- leisure time which includes the time spent in personal // Martin Paul Block care, study, religious and organizational activity, and reduces overall leisure time. Most interesting, however, is that increased shopping time also causes an increase in the amount of work-related time?) Combining the three experimental conditions men- tioned before exhibits little impact on any activity category except the experimental categories. In other words, decreases in shopping-related travel are offset by the increases in television viewing time and overall shopping time, with very little apparent effect on any other activity category. Assuming that the experimental conditions described here are a fair representation of the capability of a BCN system, then it would appear that BCN technology does not represent a major redis- tribution of the allocation of time as did television. The major impacts of BCN technology appear to be the reduction of travel time with the concomitant reduction in travel expense, and more efficient shopping behavior because of the increased time devoted to it." THE POTENTIAL IMPACT OF BROADBAND COMMUNICATION NETWORK TECHNOLOGY ON CONSUMER MARKETING COMMUNICATION: A COMPUTER SIMULATION EXPERIMENT BY Martin Paul Block A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Communication Arts and Sciences — Mass Media 1975 © Copyright by .MARTIN PAUL BLOCK 1975 ACKNOWLEDGMENTS A number of people have played a role in the completion of this dissertation and I would like to express my gratitude to them. I would first mention Dr. Gerhard J. Hanneman of the University of Southern California, who encouraged me to return to graduate school and pursue doctoral studies, and made this acknowledgment possible. I would also like to thank Dr. John P. Robinson of the University of Michigan for providing me with variance estimates from the 1965 United States Time Use Survey, making the development of the computer simulation model possible. The role of the guidance committee in the prepar- ation of a dissertation goes without saying. I think in my case the role played was an extraordinary one, requiring more than the usual expression of gratitude. Understanding that merely thanking them here is insuf- ficient, I would like to express my deepest gratitude to Dr. Thomas F. Baldwin, Dr. Gilbert D. Harrell, Dr. Thomas A. Muth, and especially my chairman, Dr. Gordon E. Miracle. ii I would also like to thank my wife, Rosemary, and daughters Bianca and Elizabeth for their patience, understanding, and numerous personal sacrifices during the preparation of this dissertation. iii TABLE OF CONTENTS Chapter I. INTRODUCTION . . . . . . . . . . Problem Statement. . . . . Technological Approach . . . Multidisciplinary Approach. . Overview of BCN Technology. . Brief History of BCN Technology . Public Policy Environment . . . Suggested Applications for BCN Tech- nology. . . . . . . . . . . Analysis Plan . . . . . . . . . II. APPLICATION OF BCN TECHNOLOGY TO MARKETING COWICAT ION . . O O O O O O 0 Marketing Communication. Communication Models. . Advertising and Promotion Retailing and Selling . Marketing Research . . Summary . . . . . . III. TECHNOLOGICAL FORECASTING METHODOLOGY. . Forecasting New Technology. . . Dialectical Methods . . . . . Teleological Methods. . . . . Experimental and Empirical Methods Analytical Methods . . . Dialectical Methods Applied to Tele- communication . . . . Empirical and Analytical Methods Applied to Telecommunication . . . . . . Computer Simulation . . . . . . . Problems in the Design of Computer Simu- lation Experiments. . . . . . . Mass Media Related Computer Simulation Models. . . . . . . . . . . iv Page {\— :6]: " 100 Chapter Page IV. TIME ALLOCATION STUDIES . . . . . . . . 117 Time as a Social Variable. . . . . . . 117 Time as a Marketing Variable. . . . . . 122 Time-Budget Studies. . . . . . . . . 127 United States Time Use Survey . . . . . 136 Analytical Models of Time-Budget Data. . . 139 Urban Planning Time-Budget Computer Simu- lation Models . . . . . . . . . . 142 V. THE SIMULATION MODEL . . . . . . . . . 150 Microanalytic Approach. . . . . . . . 150 A Description of TIMMOD . . . . . . . 157 Theoretical Probability Distributions. . . 172 Discrete Probability Distributions. . . . 178 Bernoulli Family . . . . . . . . . 178 Binomial Family . . . . . . . . . 179 Geometric Family . . . . . . . . . 180 Negative Binomial Family .' . . . . . 180 Poisson Family. . . . . . . . . . 181 Continuous Probability Distributions . . . 183 Uniform Family. . . . . . . . . . 183 Exponential Family . . . . . . . . 184 Gamma Family . . . . . . . . . . 185 Beta Family. . . . . . . . . . . 187 Normal Family . . . . . . . . . . 189 Applications to TIMMOD. . . . . . . 190 Computer Generated Random Numbers . . . . 192 Preparation of TIMMOD Input Data . . . . 195 Activity Duration Statistics. . . . . . 198 Activity Selection Statistics . . . . . 206 Sample TIMMOD Output . . . . . . . . 209 Model Validation. . . . . . . . . . 211 VI. THE SIMULATION EXPERIMENT . . . . . . . 220 Experimental Paradigm . . . . . . . . 220 Experimental and Statistical Problems in Using TIMMOD' . . . . . . . . . . 224 Analysis Method . . . . . . . . . . 227 Experimental Conditions . . . . . . . 232 Chapter Page Transportation . . . . . . . . . 232 TeleViSion . O O O O O O O O O 2 33 Marketing. . . . . . . . . . . 234 Shopping . . . . . . . . . . . 235 Limitations. . . . . . . . . . . 236 Impact of Transportation . . . . . . 238 Impact of Television. . . . . . . . 241 Impact of Shopping . . . . . . . . 244 Impact of Combined Experimental Con- ditions . . . . . . . . . . . 253 VII . SUMMARY AND CONCLUSIONS . . . . . . . 2 6 2 Summary . . . . . . . . . . . . 262 Conclusions. . . . . . . . . . . 273 Suggestions for Future Research . . . . 219 APPENDIX 0 O O O O O O O O O O O O O 2 8 2 SELECTED BIBLIOGRAPHY . . . . . . . . . . 2 8 7 vi 11. 12. 13. 14. LIST OF TABLES PENETRATION OF CABLE SYSTEMS BY SIZE OF MRKET IN 1974 O O O O O O O O 0 SUMMARY CHARACTERISTICS OF THE NATIONAL TIME USE SURVEY SAMPLE . . . . . . COMPARISON OF THE MULTINATIONAL COMPARATIVE STUDY AND TIMMOD ACTIVITY CATEGORIES. . PROVIDED ACTIVITY SELECTION STATISTICS. . ESTIMATED ACTIVITY DURATION STATISTICS. . ESTIMATED TIMMOD ACTIVITY DURATION PARA- ETERS O O O O O O O O O O O O PERCENTAGE ENGAGED IN DIFFERENT ACTIVITIES BY HOUR OF THE DAY. . . . . . . . TIMMOD REPORT . . . . . . . . . . TIMMOD SUMMARY. . . . . . . . . . COMPARISON OF THE ORIGINAL DATA AND TIMMOD OUTPUT O O O I I O O O O O O O TIMMOD RESULTS SHOWING LOW TRANSPORTATION IMPACT C O O O O O O O O O O O TIMMOD RESULTS SHOWING HIGH TRANSPORTATION IMPACT O O O O O O O O O O O O TIMMOD RESULTS SHOWING LOW TELEVISION IMPACT O O O O C C I O O O O U TIMMOD RESULTS SHOWING HIGH TELEVISION ImACT O O I C O O O O O O O 0 vii Page 18 138 197 200 203 205 208 210 212 215 239 240 242 243 Table 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. TIMMOD RESULTS SHOWING IMPACT . O O C . TIMMOD RESULTS SHOWING IMPACT. . . . . TIMMOD RESULTS SHOWING ’TIMMOD RESULTS SHOWING TIMMOD RESULTS SHOWING IMPACT.' . . . . TIMMOD RESULTS SHOWING IMPACT. . . . . TIMMOD RESULTS SHOWING LOW TELEVISION AND LOW ALL SHOPPING IMPACT. . . . . TIMMOD RESULTS SHOWING HIGH TRANSPORTATION, HIGH TELEVISION AND HIGH ALL SHOPPING IMPACT. . . . . TIMMOD RESULTS SHOWING HIGH TRANSPORTATION, NO TELEVISION AND HIGH ALL SHOPPING IMPACT. . . . . TIMMOD RESULTS SHOWING HIGH TRANSPORTATION, LOW TELEVISION AND HIGH ALL SHOPPING IMPACT. . . . . Page LOW MARKETING 0 I O O O O O O 246 HIGH MARKETING 0 O O O O O O O 247 LOW SHOPPING IMPACT . 248 HIGH SHOPPING IMPACT . 249 LOW ALL SHOPPING . . . . . . . . 251 HIGH ALL SHOPPING O O I C O C O O 252 LOW TRANSPORTATION, . O O O O O O O 255 . O O O O O O O 256 C O O O O O O O 258 O O O O O O O O 259 viii Figure 1. 11. LIST OF FIGURES Kramer Mass Media Simulation Flow Stage One: Data Disaggregation Integration . . . . . . . Kramer Mass Media Simulation Flow Stage Two: Processing Messages Reporting Exposures . . . . Brail's Activity System Simulation Model TIMMOD Main Program Flow Diagram. Diagram and Diagram and TIMMOD Setup Subroutine Flow Diagram TIMMOD Report Subroutine Flow Diagram The Uniform Distribution . . . The Exponential Distribution . . The Gamma Distribution . . . . The Beta Distribution . . . . The Normal Distribution. . . . ix Page 103 105 153 164 168 170 184 185 186 188 189 CHAPTER I INTRODUCTION Problem Statement Estimating the future impact of any communication technology is by no means an easy task. Imagine how difficult it would have been in 1935 to have estimated the impact electronic television would have on the style of life in 1965. It would have been hard to believe that electronic television would become the dominant national advertising medium, with the national general circulation magazines such as Collier's, Look, Life, and the Saturday Evening Post having either ceased publication or be struggling for survival. It would also be hard to believe that in just thirty years the watching of electronic television would occupy about 40 percent of all leisure time.1 The changes in life-style or behavior patterns in the United States during this thirty-year period can be considered in terms of changes in the allocation of time to various activities. Life-style in one important sense, according to Andreasen, can be looked upon as an allocation problem: ”given a fixed resource time, how do different groups apportion available time across 2 For example, in 1935 Sorokin and various activities?” Berger found the average adult to spent about 26 minutes listening to the radio, 29 minutes reading newspapers and magazines, and 14 minutes reading books each day.3 In 1935 the total average time consuming the mass media was about 69 minutes. In 1965, Robinson and Converse found the average time listening to the radio as a primary activity to be only 8 minutes, 84 minutes watch- ing television, 21 minutes reading newspapers and maga- zines, and 14 minutes reading books.4 In 1965 the total average time consuming the mass media was about 133 minutes, or nearly twice the 1935 average. This drama- tic change in mass media consumption is almost attribut- able entirely to television. If television viewing as a secondary activity is added the total viewing time increases to 94 minutes per day. An activity which did not exist thirty years before and accounts for one hour and a half of an average adult's day represents a dra- matic shift in the daily behavior patterns of those adults. This shift in the allocation of time is indica- tive of the profound impact television has had on society. In 1975 Broadband Communication Network (BCN) technology is in much the same situation as was tele- vision in 1935. There is no doubt that BCN technology represents in some degree the wave of the future in all human telecommunication activity. The problem is to estimate the nature and extent of the impact BCN tech- nology will have on society. If it were possible to obtain the time allocation for attending BCN events in 1985 or 1995, then it would be possible to make some estimate of the impact BCN technology has had. It is the main thrust of this work to estimate the future impact of BCN technology through the estimation of the allocation of the consumer's time as a result of the intervention of BCN technology. The tremendous communication potential of BCN technology broadens its range of application from enter- tainment and leisure to nearly every category of human activity, with perhaps the exception of night sleep. BCN technology can directly involve work, education and organization activities, semi-leisure activity such as shopping and errands, as well as leisure and entertain- ment. The range of activity is sufficiently broad to require some arbitrary limitation. Here the range of communication activity will be limited to consumer marketing communication, such as communication involved in shopping and advertising. There are two reasons for focusing the study on marketing communication. First is the popularity of marketing-related applications of BCN technology in the current literature. Normally the first applications mentioned for two-way cable television technology include pay or subscription television, in-home shopping and ticket purchasing, in-home surveillance systems such as fire and burglar alarms, and other commercially oriented applications.5 It is only recently that noncommercial applications have been considered, such as in the areas of social service delivery and urban administration.6 BCN technology has been traditionally viewed in terms of its potential commercial viability. The second reason for considering consumer marketing communications is the relative importance of such communications to society. A free market economy assumes the free flow of information about available products and services so the best choices can be made by consumers. Clearly, the range of product choice available to consumers would be severely constrained 'without marketing communications. Marketing communi- cation, including personal selling, advertising, pro- motion and related activities, is a vital part of the distribution of goods and services process in society. __t It is the primary objective of this research to predict the impact of BCN technology on consumer market- ing communications. Prediction of future impact will be made in terms of changes in life-style as indicated by the allocation of time to various activities. Although changes in life-style cross all human activity and behavior patterns, the focus will be on activity and behavior directly related to the acquisition of goods and services. A secondary objective of this research is to demonstrate a methodology for assessing the impact of changes in communication systems such as technological improvement. Technological Approach The future of BCN technology is most often dis- cussed in the popular literature from an engineering perspective with little or no consideration of the human communication system that it is supposed to augment.7 BCN technology, like most new electronic technology, is often described in a "gee whiz" manner outlining the performance capabilities of mystical electronic gadgets. Included in such discussions of BCN technology are broad bandwidth bidirectional coaxial cable systems for urban distances, satellite systems for intercity and inter- continental distances, very inexpensive digital logic and memory, user-oriented and human dominated computer terminals, and greatly increased power of data processing computers and networks of computers. Ralph Lee Smith in The Wired Nation describes the capability of cable in terms of electronic news- papers, mail service, banking and shopping facilities, electronic libraries, educational and cultural programs and community originated programs. A nation criss-crossed by telecommunication facilities with virtually unlimited voice, data, and video resources is the promise of the future. In short, writes Smith, "every home and office can contain a communication center of a breadth and flexibility to influence every aspect of private and community life."8 While Smith's vision of a future telecommunication system provides an extremely provoca- tive scenario, it nonetheless emphasizes the performance capability of the technology rather than the communication needs in the various human communication systems. In marketing terms, this approach would be characterized as being "product-oriented" rather than being "market- oriented." Reid suggests that instead of regarding tele- communication as driven by a "technological dynamic" as implied in the wired nation concept, it is possible to consider the human communication need. Technological feasibility is only a necessary, and not a sufficient condition for the adoption of a new technology in a human communication system. As Reid points out, tele- communication should be considered "not as items of electronic hardware, but as channels of human communi- cation."9 Much more is required than only examining the technology itself to understand its full potential for utilization and its ultimate impact on society. Multidisciplinary Approach Reid advocates a multidisciplinary approach emphasizing the human aspects of communication rather than the engineering performance capability of tele- communication hardware. Specifically, Reid suggests that any assessment of telecommunication technology should be based upon the following four factors: 1. What is the pattern of communication needs which telecommunications must try to satisfy? 2. What improvements to telecommunication technology are likely to be feasible, at what date, and at what cost? 3. How effective are such systems likely to be, in terms of satisfying communication need? 4. What secondary impacts would arise from the widespread introduction of such systems?10 No single academic discipline is adequate to simultaneously handle all four of these factors, requiring a multidisciplinary approach. Reid suggests a number of research areas which could have potential contribution, including information theory, applied psychology, experi- mental social psychology, management studies, urban and regional planning, geography, and sociology including content analysis, control analysis, media analysis, audience analysis, and effect analysis. To this list economics and marketing should also be added. Throughout all of these disciplines, the emphasis is on the human aspects of the communication process, and not on the technological aspects. In suggesting priorities for future research, Reid writes: The vast bulk of telecommunications research is technological in orientation. . . . This amounts to a wasteful misallocation of research resources, and some shift of effort away from technological research studies towards studies of the human aspects of telecommunications is therefore of the utmost importance.11 Departure from the technological approach in assessing new communication systems is not easy, espe- cially when the multidisciplinary implications of such a move are considered. The study of human behavior is not the same as the study of the ordered and predictable world of technology. While a multidisciplinary approach is desirable to handle the complexities of the human aspects, such an approach is often unwieldly. It is necessary to integrate various disciplines along a unified theoretical continuum to keep a multidisciplinary study manageable. Assuming that human communication involves the sending and receiving of messages through physical space, it is possible to describe the communication event in terms of the time devoted to the event. The allo- cation of time to a communication event involves a managerial decision by the participants to engage in the communication activity in lieu of other competing activity. A communication event requiring more time implies a greater communication need. Since communication occurs in physical space, a spatial dimension can also be added. Becker and Willibald define this dimension as the range, and subclassify the range in terms of local, regional, or nationwide communication.12 Examining human communi- cation need in terms of the spatial dimension is less interesting than the time dimension because of the fixed nature of physical space. It is difficult to directly manipulate physical distance. While it is possible to locate in physical space to minimize communication dis- tance, it is often impractical because of geographical barriers and Similar hindrances. The conquering of distance, as applied to telecommunication, can be interpreted as a reduction in the time required to initiate a communication event. Communication need can also be considered in terms of the cost in energy and resources that are required by the participants to maintain the communi- cation event. The more energy and resources devoted to obtaining market information by consumers for example, the greater their communication need. It is this dimension of communication need that leads to a dis- cussion of the obvious trade-off between physical movement and telecommunications.v Hanneman, for example, discusses the possibility of tele-commuting or going to work via teleconferencing rather than driving on the 13 Although, as DeSerpa argues, this kind of trade-off can be viewed as a time trade-off as well.14 freeway. 10 Time is the thread which runs through most of the considerations of human communication need. Time, and in particular models of the allocation‘and utili- zation of time, can be the theoretical linkage between the many and varied disciplines recommended by Reid in the multidisciplinary approach to the study of tele- communications. Models of the allocation and utilization of time are found in economics in the form of elasticity models, in management science in the form of time and motion studies, in urban planning in the form of planning and design models, and sociology in the study of leisure time and the problems of senior citizens. Time models occupy minor roles in the marketing literature in con- sumer behavior, and in the mass communication literature in media consumption studies. 1‘ Overview of BCN Technology A Broadband Communications Network, or BCN, is an electronically sophisticated system of high volume information linkages, such as television, between nodes in a human communication network. BCN technology describes an amalgamation of technology represented by electronic information processing equipment, com- munity antenna or cable television, and terrestial communication satellites. Although the technology is described here in terms of cable television, it is not limited to the distribution of broadband signals by 11 coaxial cable. Other more advanced technologies have been suggested, such as fiber Optics and laser communi- cation which would more efficiently perform the same function. BCN technology will be described here, how- ever, in terms of cable television because it is the most immediate and available broadband technology, and it sufficiently demonstrates the potential and function of BCN. The frequency bandwidth of standard coaxial 15 which cable is 300 MHz (million cycles per second), is much wider than a standard telephone bandwidth of 6 KHZ (thousand cycles per second), hence the reference to broadband communications. A television signal requires approximately 6 MHz of bandwidth to be transmitted, which after allowing some bandwidth for channel separation, leaves the average cable television system with an effec— tive capacity of around thirty-eight channels. Additional capacity can be obtained by adding cables. It should be mentioned that the FCC currently requires only twenty channels.16 While this bandwidth allows for many video or television signals, it also has the potential to transmit signals in both directions simul- taneously. A cable television system has the potential to become a fully two-way or interactive communication system. Cable has the potential to relieve the congestion in the broadcast spectrum and provide two-way communication. 12 This tremendous potential of cable has led some to speculate that cable communication has the potential of destroying most of the structural limitations of the present broadcast media. The increasing diversity and growth of potential cable television applications makes its reference as Community Antenna Television, or CATV, obsolete, and the BCN reference far more appropriate. The combination of urban cable systems with satellite communication systems make it technologically feasible to interconnect any number of urban systems throughout the United States and the world.17 A cable system need not be limited to the extent of its cable run. The combination of urban cable systems and computer technology provides the only feasible solution to the problem of managing return response from a large cable system.18 In addition to managing a two-way system, the addition of a computer makes it possible to bring the computing power and information storage and handling capability of a computer into the home. While one-way community antenna cable television systems have enjoyed mild success in some areas, and use of satellite in international communication is commonplace, and the computer industry continues its growth, the three basic components of BCN technology have never been successfully married in any large scaleg commercially viable demonstration. The most often cited reasons are economic for the apparent sluggish 13 introduction of BCN technology. The cable industry itself has come upon economic hard times. Baer, for example, says that much of the excitement about the cable revolution has evaporated.19 Industry over- expansion, coupled with high interest rates and pro- jections of a general economic downturn have burst the speculative bubble. Despite some current pessimism, BCN technology does have a commercially successful future as demonstrated by the number of pilot two-way systems. The current economic hardships, however, point to the deficiencies in relying on engineering performance capabilities in predicting future use of new technology. Smith describes five operating pilot two-way cable systems as of January, 1975. The systems include Bedford, Massachusetts which is a small system inter- connecting six buildings in a military engineering facility, Columbus, Ohio which is the largest of the two-way pilot systems with 718 terminals using the Coaxial Scientific Corporation system which will be discussed later, El Segundo, California using the Theta-- Com SRS syStem has 30 prototype terminals, Irving, Texas has 50 terminals placed near the headend using a standard converter with a five-button digital trans- mitter-receiver, and Princeton, New Jersey with a system designed by RCA Laboratories.20 KSmith cites —-.. - / 14 two major reasons for the relatively few number of operating pilot tests. The first is the economic problem caused by the added expense of extra equipment including return amplifiers, a computer, and home termi- nals. The second is the technical problem of suf- ficiently insulating against radio frequency interference to avoid unwanted signals in the return or upstream circuit.21 BCN technology encompasses such direct human communication activity as two-way digital return cable television and teleconferencing; and such derived human communication activity as computer-toécomputer communi- cation. BCN is different from narrowband communication networks, such as the telephone system, in its ability to efficiently transport high-volume information such as television signals to every node in a communication network. It is different from the traditional television broadcasting in the ability to transport signals point- to-point, rather than radiating signals outward indis- criminately from a central point. It Shares with both traditional broadcasting and narrowband communication networks the ability to interconnect geographically separate networks using communication satellites making the cost of communication essentially independent of distance. ‘Teleconferencing is often suggested as a means of reducing transportation costs by simulating 15 face-to-face contact through the use of television. Distant parties can communicate as though they were in the same room, without the limitation of a voice-only interaction imposed by the narrowband telephone system. Tele-offices, or going to work through television rather than the freeway, as well as substantial reduction in business trips are obvious applications of teleconfer- encing technology.22 Interactive or two-way cable television normally is used to describe a different system than indicated by teleconferencing, which assumes two-way full video sig- nals. Interactive cable television as it is most commonly described consists of a full video television signal in the outbound or downstream direction, with a digital return signal generated from a push-button response pad in the inbound or upstream direction in a single cable system. Cable television systems are constructed in a tree configuration so that signals are transmitted from an origination point, or headend, out- ward along limbs and branches. Interactive cable tele- vision is almost always described as an in-home system, whereas teleconferencing is most often described in an institutional setting because of the high cost of tele- vision origination equipment such as cameras and trans- mitters. The general model for the cable system involves only the transmission of television signals from the 16 headend to all terminals in the system simultaneously. Any channel can be viewed at any point in the system, unlike the telephone system which can direct a message to a specific point. The reason for this limitation is that cable systems are not switched, although in the future they may be.23 The cost of installing a two-way cable system was mentioned as one of the reasons for the lack of pilot two-way systems. Home terminals have been esti- mated to cost more than $500 depending upon the number of features associated with the device.24 Not only is the cost of the home terminals prohibitive, but technical problems caused by rf interference plagued all of the experiments. McVoy and Reynolds have developed a prac- tical and low cost home terminal system by modifying a standard cable converter which is required to extend the channel capacity beyond the eleven VHF channels built into a standard television receiver. With the addition of in-line code operated switches and oscil- lators each modified converter becomes a remote terminal communicating with a mini-computer. The McVoy and Reynolds system using the modified converters for terminals reduces the cost of a terminal to the $100 to $200 range.25 This estimate includes the cost of the converter itself. The problems facing two-way cable television are at least partially solved with this system. 17 Brief History of BCN Technology Cable television started in the United States in 1948 in communities that were isolated from normal tele— vision service by distance and geographical terrain.26 From the beginning as a local community antenna service, cable has grown into a major industryyngince this first system, cable has grown dramatically until in 1975 there are 3,100 operating cable systems in the United States serving nearly 6,000 communities. These cable systems currently reach a total of 8.1 million subscribers or about 12.5 percent of all television households.27 Cable has not yet reached its full potential with approximately 2,500 systems approved, but not built. ’flECHble is also missing the largest urban areas, growing primarily in smaller communities and rural areas. Cable has grown rapidly in the small towns and rural areas, and now must rely.upon expansion in the major pOpulation centers to continue its grothX/KA tabulation of markets with and without operating cable systems for the top 300 markets clearly shows the problem. Dividing the ' 1‘ a \ 2w V top 300 markets into groups of 50 by population, the top 50 have only 16 markets with operating cable systems while the last 50 markets have 40. This tabulation of the top 300 standard metropolitan statistical areas with cable information from the 1974 Broadcasting Yeapbook is shown in Table l. 18 TABLE 1 PENETRATION OF CABLE SYSTEMS BY SIZE OF MARKET IN 1974 . . Top 51- , 101— 151- 201- 251- market s1ze. so 100 150 200 250 300 Markets with no cable 34 33 19 16 15 10 Markets with cable 16 17 31 34 35 4o /" f/Future growth in the cable industry will most likely come in the larger markets where construction costs are high, competition with broadcast television keen, and government regulation more restrictive:! Early projections of cable penetration have apparently dis- counted this fact. The Sloan Commission in 1971 esti- mated that cable penetration would be between 40 and 28 60 percent by 1980. Baer now estimates that 25 percent may be too high for 1980.32/IThe problem of building cable systems in urban areas can easily be understood by simply comparing the $75,000 per mile for laying cable in the largest cities with the $4,000 per mile in the rural areas;30 To build systems in the larger cities additional services are needed to attract subscribers because of the variety of television fare available off the air. Pay cable is often discussed as a possible remedy, although it has not been an overwhelming success. In 19 1975 only about sixty systems reaching a total of 65,000 subscribers offered any pay cable.31 Despite the apparent gloom, most experts agree that cable will continue to grow, but at a rate slower than anticipated at the beginning of the decade. Price summarizes the.— consensus by writing: We are left, then, with a cable technology which, all agree, will grow at medium speed during the next decade. It is almost certain that the innovation will mean a richer choice in enter- tainment offerings on television for millions of Americans. To date, almost all the considerations of the problems of the cable television industry have assumed that cable is a one-way communications medium. Two-way services, which could drastically alter the competitive position of cable in the larger metropolitan areas, have not been developed. The reason for the lack of develop- ment of two-way services is what Baer terms the "chicken and egg" barrier. The cable industry built on relatively low monthly subscriber fees does not see sufficient revenues from two-way services to justify the large investment. On the other hand, two-way services cannot be developed until the investment has been made.33_, Until recently, demonstrations of two-way cable have been limited to very small equipment demonstrations. In 1974, the National Science Foundation funded a number of developmental projects to design experiments in the area of urban administration and social service delivery 20 using two-way cable television technology. Michigan State University was the recipient of one of the planning grants to work with the cable system in Rockford, Illinois. The Rockford system was selected because of the generalizable demographic characteristics of the Rockford community, and the technical excellence of the cable system itself. The Michigan State University research project was conducted in two phases, a design phase followed by an implementation phase commencing in mid- 1975. Among the proposed experiments are a program to train firefighters to upgrade their skill using inter- active training material as well as simulations, a pro- gram to diagnose developmental delays in children under age 5 in the home, a large-scale cable information and referral service based upon a series of interactive television vignettes describing various social services and programs within the community, providing an extension and supplement to elementary science education through teleconferencing and computer-aided instruction, and a series of legal communication applications including an automated legal library, publication of court-generated information, and the taking of depositions. The five proposed experiments involve both teleconferencing and digital return interactive television with a number of institutions in the Rockford community, and hundreds of private homes.34 21 The National Science Foundation has funded Michigan State University to conduct the firefighter training experiment along with experiments being con- ducted by New York University and the Rand Corporation. In total the National Science Foundation has funded nearly $2 million, which should have some impact on the continued development of cable technology.35 Public Poligy Environment In addition to the economic problems incumbent upon BCN technology is an excessive system of public regulation. According to Branscomb the present rules regulating cable television have saddled a new technology with too much reSponSibility in its infancy before it has had time to develop the economic base necessary to sustain such public commitments.36 According to Brans- comb, the cable industry is regulated by a three-tiered system, including local, state, and federal regulation. A cable operator must meet FCC requirements, conform to any state regulation that may exist, and negotiate a franchise with the local government unit.37 Since then the FCC has established very complex rules for what can and cannot be carried by a cable system determined by an intricate set of procedures and formulas in the FCC Cable Television Report and Order 38 issued in 1972. Sardella describes the 1972 rules for the top 100 markets as restricting the importation 22 of distant signals to two channels, requiring channel capacity of at least twenty channels, the provision of three free channels including a channel for public access, local government and educational uses, and pro— vision for mandatory channel capacity tied to usage, allowances for channel leasing, and rules governing programs provided for additional charge or pay TV. The 1972 rules also require these cable systems to maintain a system having the technical capacity for nonvoice return communication.39/?The complexity of the rules makes it difficult for the average cable operator to operate without expensive legal resources. It is dif- ficult for single system operations to survive without the economies obtained in pooling systems and legal resources.é?, Not only is the regulation that has been imposed upon cable television, as described by Branscomb,41 excessive but often favors other competing media. An example of this can be seen in the FCC pay television rules. In the struggling cable industry, pay television is often described as the means of providing the financial basis to help develop cable systems, especially in the large urban areas. According to Baer and Pilnick most cable operators consider pay cable programming the service that is most likely to make their systems more attractive and therefore more economically feasible.42 23 Pay cable, however, is regulated by the FCC's "anti- siphoning” rules which attempt to minimize any compe- titive threat to commercial broadcast television and film distribution industries. The National Association of Broadcasters has been very successful in its lobby against pay television of any kind. Some of the rather severe restrictions placed on pay cable include a ban on all feature films which have a release date more than two years old and less than ten years old. It should be pointed out that films between two and ten years old constitute the bulk of broadcast film libraries. Other restrictions include a ban on sports events that have been seen on broadcast television at any time in the prior two years, and a ban on any series-type programs.43 While the sports restriction has been recently eased, the rules for the most part have constituted one more barrier for the struggling technology. The economic and regulatory barriers for cable television, however, do not necessarily negate the future success of the technology. The barriers may only delay the development and make it difficult to predict when various milestones will be passed. Suggested Applications for BCN Technology The pioneer article describing how the new BCN technology could revolutionize the entire process of 24 the distribution of goods and services was written by Baran in 1967. Baran considered the feasibility of the development of various aspects of "computer-communications" technology and the resulting nature of the pattern of changes that may occur by the end of this century. The potential impact of BCN technology on marketing is emphasized when Baran writes: Our concept of market segmentation is highly depen- dent upon information flow structures. Change the information flow mechanism by providing instan- taneous feedback to the manufacturer and the entire market segmentation process must change. 44 Baran goes on to describe marketing research, retailing and merchandising, and advertising examples of BCN technology applications. Bogart in 1973 makes the point that speculation about the future of advertising requires first a consid- eration of the developments that might occur in the mass media. The mass media, according to Bogart, are in the midst of a "technological revolution which is bound to accelerate in the rest of this century."45 Changes in the media inevitably carry implications for adver- tising, buying habits, and retailing practice. Some of the implications for advertising created 5 by BCN technology have been suggested elsewhere by this E author. The first of these include improved research 2 and evaluation methodology with the introduction of automated advertising evaluation systems that will 25 include instantaneous audience measurement with associated purchase behavior measurement. A second area might be an entirely new communication medium, such as a formal consumer forum. Such forums might be either publically or privately controlled and could become a major com- munication linkage between the consumer and the marketer. A third, and perhaps obvious, implication for advertising will be the introduction of an entirely new advertising message format. Commercials and sales messages may begin to look more like a kind of electronic catalog with ordering or response capability integrated into the pro- duction. A fourth implication, partially generated by the other three, will be the increased emphasis on tele- vision production, especially at low cost to accommodate local retailers.46 Dordick and Hanneman point to the need for the application of BCN technology to provide the better and quicker market information needed in the future. BCN technology would mean more information obtained in real-time, using nonlabor intensive methods, about smaller and better specified population segments.47 In addition to these few articles dealing directly with the application of BCN technology to marketing, there have been numerous articles about BCN technology and two-way cable television which have compiled lists of applications for the technology. Smith, for example, lists home library services, facsimile data services, 26 delivery of mail, crime detection and prevention, and travel as the major application areas for BCN tech- nology.4 Travel includes such activities as business and shopping where the communication technology can be used as a substitute for travel. The National Academy of Engineering lists thirteen potential services for BCN technology: 1. ‘0 m \1 m o 10. Give the subscriber the right to select any one of twelve or more free television channels as the demand develops, including local TV stations, distant TV broadcast signals, locally originated nonbroadcast programming, and nationally distributed cable programming via domestic satellite systems. Sufficient band- width exists to provide for cultural, edu- cational, ethnically oriented, and religious channels, as well as entertainment. Provide restricted channels that would be available only to an authorized group such as doctors, etc. Provide the means to order merchandise from product demonstrations on available channels. Permit viewer participation in public preference polling, with optional means for protecting the identity of the reSponder. Provide warning of fire, medical or intrusion emergencies in the home. This information could be forwarded directly to the municipal emergency service authorities. Provide educational channels with a student response capacity. Provide statistical data relating to television viewing preferences. Provide turn-on service in the home for lights, heat, warning systems, etc. Provide readings of the various meters used for gas, electricity, and water, send the reading directly into the utility company computer, and return the billing to the user. Provide facsimile page-type or strip-type hardcopy or electronic readouts in the sub- scriber's home for messages or statements. 27 11. Provide means for capturing and storing con- tinuous display single pictures from a shared channel (frame grabber) as a means of delivering slide information into the home. 12. Provide channels on a lease or free basis for use in programming by independent persons or agencies. 13. Provide access to premium programming with color and resolution capacities superior to broadcast standards.49 Baer has prepared a compilation of proposed services for BCN technology from various reports, FCC filings, cor- porate brochures, and advertising materials.50 Included in the list are subscription television, remote shopping, electronic delivery of newspapers and periodicals, catalog displays, ticket sales, banking services, local auction sales and swap shops, television ratings, and market research surveys.51 Baran conducted a Delphi Process study in 1970 to obtain estimates of the service characteristics of the thirty interactive cable-related home services. Overall, Baran found the new industry of electronic services to the home is estimated as essentially start- ing in 1980 and building to a twenty billion dollar industry.52 Shopping transactions or store catalogs which describe or Show goods at the request of the buyer were predicted as early as 1977 and as late as 1990. Other marketing related services such as mass mail and direct mail advertising are predicted as early as 1930.53 28 The list of potential applications for BCN technology spans a broad range of activity in both public service and commercial areas. The list includes services that cover the entire spectrum of economic feasibility. It is obvious that the list was developed heavily emphasizing the performance characteristics of the electronic hardwarea' This research is an attempt to depart from the more common technological approach and employ the multidisciplinary approach advocated by Reid. Analysis Plan Computer simulation used as a research methodology fits well with the multidisciplinary approach. Computer simulation allows the investigator to be eclectic in his approach and combine many different disciplines and methodologies to develop the complex and well-Specified models necessary. Computer simulation is also an ideal forecasting methodology because it allows vicarious experimentation. Computer simulation is one of the few research approaches that allow asking fanciful "what if” questions. A hypothetical world having the advantage of a new communication technology can be created using the electronic symbol manipulation capability of a digital computer with much less effort and expense than in the real world. 29 The first step in this research will be to analyze the existing communication patterns that exist in the marketing process from the point-of—view of the consumer. This will be the topic of Chapter II which immediately follows this introduction. Chapter III will deal with the problems of tech- nology forecasting and a review of technology forecasting methodology. Also included in this chapter will be a number of examples of telecommunication related techno- logical forecasts, and a detailed examination of computer simulation as a research methodology. Chapter IV will explain the basis of the simu- lation model which is human activity analysis. The chapter will begin with a review of the theoretical positions of time as a social variable and a review of time-budget studies. The 1965 United States Time Use Survey conducted by the Institute for Social Research at the University of Michigan will be outlined in some detail because it provides the essential data base for the simulation model. A discussion of simulation and process models based upon human activity data will follow, with Special emphasis on urban planning models. The development of computer Simulation time model (TIMMOD) itself will be described in Chapter V. The chapter will begin with a functional description of the model and rationale for its development. This will 30 be followed by a discussion of theoretical probability distributions because of their crucial role in the operation of the model. An explanation of the derivation and transformation of the relevant input data for the model from the published results of the 1965 United States Time Use Survey, and a discussion of the problem of verification of the model will end the chapter. The simulation experiment which will evaluate the potential impact of BCN technology will be described in Chapter VI. This chapter will begin with a discussion of the importance of the allocation of time and a description of the experimental paradigm used in the research. A discussion of some of the problems and limitations of computer simulation experiments comes next. The treatment conditions, including measures for communication/transportation tradeoff, programming variety, and availability of both marketing and non- marketing services, will precede the actual simulation runs. The simulation runs will then be discussed and interpreted. The implication of different levels of cable penetration will be discussed to end the chapter. Chapter VII is the final chapter and will con- tain the summary, conclusions, and recommendations for additional research. In appendices will be the FORTRAN source coding for the TIMMOD main program and all the supporting subroutines. CHAPTE R I- -NOTES 1John P. Robinson and Philip E. Converse, "The Impact of Television on Mass Media Usages: A Cross- National Comparison," in The Use of Time, ed. A. Szalai (The Hague: Mouton, 1972), p. 211. 2Alan R. Andreasen, "Leisure, Mobility, and Life-Style Patterns," in Changipg Marketing Systems, ed. R. Moyer (Chicago: American Marketing Association, 1967), p. 55. 3Pitirim A. Sorokin and Clarence Q. Berger, Time-Budgets of Human Behavior (Cambridge: Harvard University Press, 1939), p. 62. 4Robinson and Converse, "The Impact of Television on Mass Media Usages," p. 199. 5Walter S. Baer, Interactive Television: Prospects for Two-Way Services on CableCTSanta Monica, Calif.: Randi Corporation, 1971). 6Thomas F. Baldwin, Bradley S. Greenberg, and Thomas A. Muth, Design Study for Urban Telecommunication Experiments (East Lansing, Michi: Michigan State Uni- versity, I974). 7Alex Reid, New Directions in Telecommunications Research, A report prepared for the Sloan Commission on CEEIe Communications, 1971, p. 3. 8Ralph Lee Smith, The Wired Nation (New York: Harper & Row, 1972), p. 2. 9Reid, New Directions in Telecommunications Research, p. 3. 31 32 loIbid. 11Ibid. 12Dietrich Becker and Gunther E. Willibald, "Classification and Assessment of Telecommunication Services in Broad-Band Networks," IEEE Transactions on Communications COM-23 (January 1975): 64. 13Gerhard J. Hanneman, "Communication Substitutes for Transportation: Movement of Information Versus the Movement of People," University of Southern California, Los Angeles, California, 1974, p. l. (Mimeographed.) 14A. C. DeSerpa, "A Theory of the Economics of Time," The Economic Journal 81 (December 1971): 828. 15Delmer C. Ports, "Trends in Cable TV," IEEE Transactions on Communications COM-23 (January 19755: 95. 16Vincent Sardella, "Interactive Cable: Federal Policy and Programs," IEEE Transactions on Communications COM-23 (January 1975): 172. 17Ports, "Trends in Cable TV," p. 96. lerid., p. 95. 19Walter S. Baer, "Cable Television in the United States--Revolution or Evolution?" Santa Monica, Cali- fornia, Rand Corporation, 1974, p. l. (Mimeographed.) 20Ernest K. Smith, "Pilot Two-Way CATV Systems," IEEE Transactions on Communications COM-23 (January 19755: 113-14. 211bid., p. 111. 22Hanneman,"Communication Substitutes for Transportation," p. 16. 23Kenneth Rose and Richard K. Stevens, "Design of a Switched Broad-Band Communications Network for Interactive Services," IEEE Transaction on Communications COM-23 (January 1975): 49355. 33 24Ibid., p. 54. 25D. Stevens McVoy and Richard Reynolds, "Cost Barrier Cracked in Two-Way Cable TV," Electronics (February 20, 1975), p. 100. 26Ports, "Trends in Cable TV," p. 92. 27Broadcasting Cable Sourcebook 1975, p. 5. 280n the Cable: The Television of Abundance, Report of the Sloan Commission on Cable Communications (New York: McGraw-Hill, 1971), p. 39. 29Baer, Cable Television in the United States, p. 4. 30Broadcastinngable Sourcebook 1975, p. 5. 3lIbid. 32 Monroe E. Price, "The Illusions of Cable Tele- vision," Journal of Communication 24 (Summer 1974): 76. 33Baer, Cable Television in the United States, p. 8. 34Thomas F. Baldwin, Bradley S. Greenberg, and Thomas A. Muth, Experimental Applications of Two-Way Eagle Communications in Urban Administration and Social Serviee Delivery, MiChigan State University, East Lansing, Michigan, 1975. 35"Projects Will Test 2-Way Cable TV for Future Uses," New York Times, August 3, 1975. 36Anne W. Branscomb, "The Cable Fable: Will It Come True?" Journal of Communication 25 (Winter 1975): 50. 37Ibid. 38Steven R. Rivkin, Cable Television: A Guide to Federal Regulations (Santa Monica, Calif.: Rand Corporation, 1973). PP. 93-141. 34 39 Sardella, "Interactive Cable,‘ p. 172. 4OBranscomb, ”The Cable Fable," p. 54. 41Ibid., p. 50. 42Walter S. Baer and Carl Pilnick, Pay Tele- vision at the Crossroads (Santa Monica, Calif.: Rand Corporation, 1974). p. 2. 43Ibidol pp. 6-7. 44Paul Baran, "Some Changes in Information Technology Affecting Marketing in the Year 2000," in Changing Marketing §ystems, ed. R. Moyer (Chicago: American Marketing AssoCiation, 1967), p. 76. 45Leo Bogart, "As Media Change, How Will Adver- tising?" Journal of Advertisipquesearch 13 (October 1973): 25. 46Martin Block, "Two-Way Advertising: An Appli- cation of Broadband Communication Network Technology," East Lansing, Michigan, Michigan State University, 1975, pp. 12-13. (Mimeographed.) 47Herbert S. Dordick and Gerhard J. Hanneman, "Utilization of Broadband Communications Network Tech- nology for Proprietary Market Research," Los Angeles, California, University of Southern California, 1975, p. 9. (Mimeographed.) 48Smith, The Wired Nation, pp. 86-88. 49Communication Technology for Urban Improvement, Report to the Department of Housing and’Urban Develop- ment (Washington, D.C.: National Academy of Engineering, 1971). pp. 21-22. 50Baer, Interactive Television. SlIbid., p. 12. 35 52Paul Baran, "Broad-Band Interactive Communi- cation Services to the Home: Part I--Potential Market Demand," IEEE Transactions on Communications COM—23 (January 1975): 5-15. 53Ibid., p. 9. CHAPTER II APPLICATION OF BCN TECHNOLOGY TO MARKETING COMMUNICATION Marketing Communication A number of traditional marketing activities can be described under the rubric of marketing communication, including personal selling, advertising, packaging, point-of-purchase, direct mail, product sampling, pub- licity, and public relations. The use of the term "marketing communication" implies a consolidation of the communication activity involved in the marketing. effort of the firm, which until recently, as Ray points out, has not been seriously attempted.l Stidsen and Schutte describe marketing itself as a communication process rather than as it is traditionally described as a control process. Marketing as a communi- cation process is considered as a dialogue between pro- ducers and consumers, whereas marketing as a control process is seen as a well-directed and efficiently implemented producer monologue. The marketing communi- cation process is described by Stidsen and Schutte in terms of four strategic dimensions: 36 37 l. The boundaries of marketing concern and responsi- bility, such as size and dimensionality of the marketing "scene." 2. The set of potential outcomes of a particular marketing effort or of marketing in general given the products and consumers involved. 3. The set of available and potentially available methods of communication between producers and consumers. 4. The division of tasks among various marketing agents (including producers and consumers), and the relative degree of control available to each agent or group of agents. It is conceptually appealing to consider marketing as a communication process, rather than as a control process. If marketing is to be considered as a communication pro- cess, then the focus of this study should be on the methods of marketing communication, or the communication between producers and consumers. The methods of marketing communication are represented by a variety of communi- cation media in a variety of communication Situations. For example, marketing communication occurs face-to-face, over the telephone, through the mail, and the mass media. It can occur in Situations before, during, and after a purchase or transaction. Marketing communication methods can be classified in a variety of ways, by function, as is normally the case in the marketing literature, or by elements of the communication situation, as is the case in most behavioral science literature. Classifying methods of marketing communication according to their function provides cate- gories such as advertising and promotion, retailing and 38 selling, and market research. The behavioral science approach typically analyzes the communication situation in terms of a source, message, channel, and receiver. Criteria such as the direction of communication flow, that is the communication being primarily in the producer to consumer direction, about the same, or in the opposite direction; the initiation of the communication, that is either producer or consumer; and the communication channel, such as face-to-face, mediated faceFto-face, and mass media. Advertising would be defined as market- ing communication that is producer initiated, flowing primarily in the producer to consumer direction through the mass media. Retailing as a communication process would be mainly face-to-face (ignoring direct mail) consumer initiated communication flowing primarily in the producer to consumer direction. Market research as a communication process would be for the most part pro- ducer initiated mediated face-to-face (telephone or mail) communication flowing primarily in the consumer to producer direction. It is interesting to note that almost all of the methods of marketing communication are oriented in the producer to consumer direction. Market research, while characterized as being primarily in the consumer to pro- ducer direction, is producer controlled and very low volume when compared to advertising and retailing. 39 Despite the emphasis on a dialogue between producers and consumers in the marketing concept, most marketing communication is clearly a producer monologue. Stidsen and Schutte describe the problem as follows: Perhaps the basic reason for the paucity of consumer-producer communication channels is that marketing is still viewed as selling, the rhetoric of the marketing concept notwithstanding. The necessary technology clearly exists (e.g., tele- phones, two-way cables, and electronic processing equipment), but the economic rationale for and commitment to giving the consumer a voice in producer decisions is clearly lacking.3 The potential impact of BCN technology goes to the very heart of the marketing concept. Bogart characterizes the mass media as being in the midst of a technological revolution which is bound to accelerate in the rest of this century. Mentioned among other communication technology is two-way cable that will make possible the direct ordering of advertised goods and the display of on-demand information and entertainment. The impact on marketing, and in par- ticular advertising is profound. Bogart writes: Changes in media thus inevitably carry implications for the content and targeting of advertising mes- sages. These changes parallel changes in buying habits and in retailing practice. Bogart believes that technological change in the mass media will have the effect of reducing the mass audience and increasing the number of consumer options. Ray describes telecommunication technology as the most important "future breakthrough" which will 40 change the very nature of marketing communication. About the impact of telecommunication technology on marketing communication, Ray writes: Probably the clearest trend related to marketing communication is that the hardware is changing dramatically. Cable TV, satellite communications, videOphones, computer-aided instruction, videotape, holography-~all of these and more promise not only to change the media but also to change control of the media. With more channels and home and office information centers, the control of communication will shift from the sender Side to the receiver side. Mass media communications will resemble personal selling, and "personal" selling will be conducted through the media to create structural similarity among the elements of the communication mix than currently exists.5 Although, not all those who speculate on the future of marketing communication agree with Ray. Banks reports a survey of marketing and advertising executives about the past, current, and future roles of advertising and promotion. Notably absent from the list of future trends is any reference to telecommunication technology.6 This is akin to speculating about the future of advertis- ing and promotion in 1935 without any mention of the new medium television. Communication Models To develop a typology of marketing communication methods, a brief review of communication models is necessary. In 1952, Deutsch characterized two basic types of communication models; flow models such as those suggested by Lashley in psychohydraulics, and 41 cybernetic models as defined by Wiener.7 This funda- mental dichotomy has remained basically unchanged through today. Johnson and Klare speak of "diagrammatic" models and "mathematic" models of human communication.8 Diagrammatic models are very common in communi— cation research. The most common type of diagrammatic model is a relatively simple flow model showing the transfer of information from a "source" to a "receiver" in the communication dyad. Osgood, for example, has suggested a model which begins with a source, which encodes a message, which is sent to a receiver who decodes the message. Schramm has modified this basic model so as to include feedback. Westley and MacLean have modified the basic model even more to accommodate mass communication. The Westley and MacLean model complicates the simple dyad model to allow for fortuitous communication and feedback since orientative feedback is not directly in the mass communication situation, as it is in the interpersonal situation. This mass com- munication model also builds in the role of reporter, and treats this function as an extension of the environ- ment.9 An interesting extension of the diagrammatic model is the geometric treatment of communication 10 "networks" as was first suggested by Bavelas in 1948. Although network analysis has progressed substantially 42 since then, its application is primarily in the study of social organization and not appropriate here. The basic mathematical communication model was developed by Shannon and Weaver in 1949. While this model has all the elements of the diagrammatic models, such as the "information" source which produces a message which is "transmitted" over a "channel" to a "receiver" and "destination," the Shannon and Weaver model also includes the concept of noise.11 I The concept of noise is a concept central to information theory and can be described in pure mathe- matical terms. The information theorist considers entropy and redundancy as opposite ends of a continuum, and can define entropy or uncertainty using a simple mathematical formula: Hmax = logzn The expression Hm is the maximum amount of entropy in ax the system and n is the number of equally probable out- comes. Similarly, there are mathematical equations which describe noise as a function of bandwidth and power of transmission. Information theory is replete which a variety of interesting concepts such as coupling, networks, and channel capacity. While information theory and mathematical models of communication are interesting, they have only limited 43 application in the development of a typology of market- ing communication methods. The primary reason for this is simply the lack of empirical data describing different marketing communication situations. In order to develop a typology, then, it is necessary to rely on simple diagrammatic models to describe the communication situ- ation with such basic elements as a source, receiver, and feedback. A typology of marketing communication methods can be described in terms of the elements of diagrammatic communication models. Three basic types of marketing communication can be easily identified in these terms that generally correspond to corporate functions. The first type of marketing communication is advertising and promotion. Advertising involVes the sending of messages~from3the advertiser source to the consumer audience receiver through the mass media communication channel. This type of marketing communication does not involve any immediate feedback and is normally charac- terized as one-way communication. Advertising and pro- motion are almoSt completely dependent upon the traditional mass media. Itis also one form of marketing communi- cation that is passively consumed. A secénd type of marketing communication is retailing and selling. Retailing and selling is a ‘:;"“1? I o I separate category because 1t 1nvolves act1ve 44 participation on the part of consumers, generally does not rely on the mass media, and often requires trans- portation of either the consumer himself or the goods that are purchased. Transportation is not typically involved in advertising, although the function of retail- ing and advertising on the surface may appear to be the same. Retailing and selling involve the direct exchange of both goods and information, and either involve face- to-face interaction between the consumer and the seller or a substitute for face-to-face communication. The third type of marketing communication is marketing research.= It is distinct from advertising and retailing in its emphasis on feedback from the con- sumer to the producer. Marketing research is the only formal communication method in most firms for consumer feedback except sales. It should be pointed out that sales are not really communication in that they repre- sent only choice among a set of constrained alternatives. As mentioned before, marketing research provides only a small amount of return communication. Certainly marketing research cannot hope to create any dialogue between producers and consumers. Also, the producer has complete control over the format and content of the information and is likely to emphasize issues important to producers rather than issues that are important to consumers. No doubt the lack of a more adequate feedback 45 channel from consumers to producers is a contributing factor in much of the consumerist criticism of marketing. An analysis of the potential impact of a new communication technology such as BCN technology on these types of marketing communication must begin with the communication as it currently exists. Even though a major impact of the technology may be to dissolve the definition between the types of marketing communi- cation such as advertising and personal selling. Potential impacts of BCN technology, then, will be described in terms of each of the three types of marketing communi- cation discussed here. A consolidation of these impacts will be considered later in Chapter VI. Advertising and Promotion Advertising and promotion are primarily mass media communication activities. To describe the impact of BCN technology on advertising and promotion it is necessary to consider the impact on the mass media. As Bogart writes, "to speculate about the future of advertising requires that we first consider some of the developments that lie ahead for the mass media with which most adver- tising has always been linked."12 Some of the implications of BCN technology for advertising and the mass media have already been sug- gested. There is little doubt that BCN technology is growing and is destined to become a very important 46 communications medium, and that the traditional one-way mass media, including both print and broadcast, will become highly vulnerable to the increased penetration of first multi-channel systems, then networks, then inter- active television. Also of major importance is that the public interest in interactive television is strong and no doubt a portion of the medium will be devoted to public interest programming.l3 Although the additional channel capacity will easily accommodate commercial programming, public interest programming offered through a BCN may for the first time provide real competition with commercial programming. The public interest in BCN technology is demonstrated by the regulation dis- cussed in Chapter I. The introduction of BCN technology as an adver- tising medium will serve to further blur the distinction between advertising and retailing because of the con- sumer's opportunity to immediately respond. It will greatly increase the efficiency of the advertiser's ability to provide real prospects with little waste. It will involve a whole new message format, namely, interactive television. Speculation about the future of advertising as a result of the impact of BCN technology might run as follows. First result of the technology would be the introduction of an automated advertising evaluation 47 system aimed at traditional one-way broadcast television. The evaluation system will adopt newer message formats as the technology develops, and provides instantaneous audience measurement with associated purchase behavior measurement. It is a simple matter to include monitoring capability in a BCN system that can provide second by second monitoring of an in-home television set. Unlike conventional television ratings calculated for fifteen minute time segments, second by second ratings are possible. It is possible to collect product purchase behavior by administering a questionnaire on a continual basis with the push-button return for response which would be directly recorded by computer as described in Chapter I. It would also be possible to conduct copy tests having respondents viewing commercials in their homes, reacting to the commercials using the response pad. A second result of BCN technology might be the establishment of consumer forums. Such forums might be either publically or privately organized and could potentially become the major communication link between consumer and producer to make the producer-consumer dialogue in the communication process conception of marketing a reality. Consumer forums could involve a combination of teleconferencing and digital return interactive television, and might be a governmental 48 service Sponsored by consumer affairs agencies, or a trade association, or be a private corporation. A third result might be the introduction of an entirely new advertising message format, namely two-way television. Commercials and sales messages will begin to look more like a kind of electronic catalog with ordering or response capability integrated into the program. Commercials will no longer be the Short repetitious messages as they are now, but probably longer and more detailed messages that would be sought by the consumer in much the same way as a catalog is used. A fourth, and very obvious, result will be the increased emphasis on television production, especially at low cost to accommodate the local advertiser. It will also be necessary for future television production personnel to be familiar with computer techniques as well as traditional television production to effectively produce future messages using interactive television. While speculation such as this may be interesting, it does not provide much insight in the problem of assessing the future impact of BCN technology. When considering the impact of BCN technology on advertising and promotion, it is clear that the discussion will center on television advertising, with the impact on other com- peting advertising media considered as secondary. A BCN system provides the potential for both increased diversity 49 and attractiveness of television program material. Increased diversity is made possible through greatly expanded channel capacity as discussed before. Attrac- tiveness of the medium Should be greatly enhanced by the increased diversity as well as the opportunity to provide interactive program material. Unfortunately there is no direct way of deter- mining the effect of increased program diversity and interactive programming. The best substitute is the experience of one-way cable television which can provide some indication of the effect of increased program diversity. Agostino did a study of television consumption behavior in terms of channel use between cable sub- scribers and broadcast viewers. The study consists of a comparison of thirty-one cable households and thirty- one noncable households in several American Research Bureau markets. The comparisons are made on the basis of ARB diary reports. The study, although based on a very small sample, has a number of interesting findings. Agostino summarizes the major findings as follows: 1. individual television viewers whether by broad- cast or cable generally watch only three to five channels; 2. use of channels increases only slightly as the number of available channels increases; 3. cable viewers in three of the five markets utilized more channels than broadcast viewers; 4. cable subscribers view more prime time tele- vision than broadcast viewers; , ,LI n‘v 2| a v.. nut 50 5. heavy television consumers utilize more channels and distribute their viewing more evenly across channels than low consumers: 6. household viewing groups generally use only three to six channels; 7. households with children utilize more channels and distribute channel use more evenly than households without children; 8. market differences have greater influence on the number of channels used, than differences of reception type; 9. though some differences among viewing groups have been noted, all viewers tightly concentrate their viewing time on the local network-affiliated stations; 10. the expanded channel choice offered by cable service does not alter this concentration on network programming: 11. but does alter the relative popularity of the local network stations.l Overall, Agostino found relatively little impact of tele- vision consumption behavior as a result of cable service. Prime time viewing of network programming accounts for the majority of the behavior. Such homogeneity as led Robinson to describe television's role as "electronic innkeeper to the less active minds of society."15 Robinson also believes that the best prediction of future use of television is "status quo." There seems to be a predictable limit to the amount of time people will spend watching television, and that in the near future television in the United States may be approaching its maximum audience, simply because nearly every household has at least one television set.16 In general it appears that one-way television, whether it is cable or not, will continue to be consumed in approximately the same way as it is today. 51 There is one difference, however, that might be overlooked because of the small sample size in the Agostino study. Agostino reports the mean total viewing time in number of fifteen-minute units per week for individuals with cable as 101.34, and without cable as 90.86.17 Cable subscribers appear to spend over 10 percent more time per week watching television than noncable subscribers. However, Agostino does not report this result as being a statistically significant difference. However, some mention should be made of the statistical test applied by Agostino as being a two factor treatment by level analysis of variance. The variability across markets, ranging from 70.68 to 110.81 for individuals with cable for example, is sta- tistically significant. Treating different ARB markets along with the type of television reception would seem less important if the number of markets were increased beyond five. Also, even if the difference between those with cable and without cable were significant, many other factors could account for the difference than merely the availability of cable. For example, those more interested in viewing television would be more likely to be cable subscribers. With no empirical evidence available, it seems most reasonable that households spend more time watching television as a result of BCN technology. It will be 52 assumed here that making the cable system two-way and providing interactive programming would make watching television even more attractive. In the absence of any contradictory evidence the overall time spent viewing television per household would increase as a result of the application BCN technology. Without BCN technology, the time spent viewing television would remain approxi- mately the same. The increase in television viewing time, then, will be considered as the major impact of BCN technology on advertising and promotion. Retailing and Selling Retailing and selling has been described here as a type of marketing communication involving either face-to-face communication or some substitute for face- to-face communication. Retailing and selling, or shopping, clearly involves some form of transportation, either the consumer must go to the seller or vice versa. For the most part, shopping implies some kind of physical exchange of goods and services. The communication substitute for some of the transportation in the shopping process is an obvious result of the application of BCN technology to the process. Clearly much of the transportation involved in the shopping process does not involve the physical movement of goods, but rather moving in Space to obtain information to make a purchase decision. Examples are 53 comparison shopping or window shopping. The use of catalogs and the telephone in shopping illustrate the need for information. Although, as Tauber warns, the shopping process offers other benefits than simple exposure to products.17 Tauber suggests a number of motivations for shopping other than obtaining simple product information, including personal motives such as role playing, diversion, self-gratification, learning about new trends, physical activity, and sensory stimu- lation; and social motives such as social experiences outside the home, communication with others having a Similar interest, peer group attraCtion, status and authority, and the pleasure of bargaining.18 Tauber believes that retailers should do more than emphasize the promotion and distribution of goods as they tradi- tionally have, and should consider themselves as being part of the social-recreational industry. As Tauber points out: As businesses which offer social and recreational appeal, retailers must acknowledge that they are competing directly for the consumer's time and money with other alternatives that provide similar benefits.1 However, neither Tauber's nor any other research indi- cates which of the shopping motives are the most impor- tant. Since shepping with a BCN system is essentially an in-home activity, it is worthwhile to examine some 54 of the characteristics of present-day urban in-home shoppers. Gillett cites estimates that in-home sales accounted for 9 percent of total general merchandise sales in 1970 and could reach 11 percent by 1975. Catalog sales have increased at a rate almost twice that of all general merchandise sales.20 Shopping con- venience and the reduction of shopping time and effort are the usual reasons given for the increase in in-home shOpping. Gillett in a study of 210 Grand Rapids homemakers found that the in-home shopper is not a "captive" market, but are better described as a "modern" shopper. Women buying at home are also active store shoppers and do not find store shOpping difficult or unpleasant. In- home Shoppers are convenience oriented and flexible in their choice of shopping behavior. Also important, is that in-home shoppers are more affluent and better educated than other Shoppers. Those shoppers that dis- like in-home shopping tend to buy little at home, are lower socio-economic status and are somewhat older.21 Gillett's study includes only telephone and mail shopping, not door-to-door solicitation. Anderson identified convenience-oriented consumer using data from a consumer panel reporting convenience food consumption. Anderson supports Gillett's conclusion describing the convenience-oriented shopper as enjoying 55 higher socio-economic status.22 Reynolds did a mail survey of female homemakers in Georgia to determine factors that affect catalog buying behavior. Reynolds found that current catalog buyers do not perceive cur- rent shopping conditions as either too frustrating or unentertaining, but the Opposite. Reynolds concludes: "Barring a continued major problem in mobility as a result of gasoline shortages and assuming a continuing thrust by merchants to provide enjoyable shopping environments, convenience is not likely to regain a major position as a determinant of catalog buying."23 Rather, Reynolds found that the catalog shOpper is most influenced by the product offering and selection found in the catalog.24 Not all in-home Shoppers are more affluent and better educated as was found by Peters and Ford in a study of in-home cosmetic customers in Wisconsin. Peters and Ford found that the in-home Shopper has less access to a car, tends to be less educated, has lower income, and is less likely to be in a household with a profes- sional household head.25 Peters and Ford, however, limit their consideration of the in-home shopper to the door-to-door solicitation. Other forms of in-home shopping were not considered. The difference in the socio-economic status of in-home or the "modern" shopper and the "traditional" 56 shopper points to one of the most profound potential impacts of BCN technology on consumer marketing com- munication. As Thorelli points out, there tends to be a greater concentration of market information among the more educated and higher income groups in the world.26 The more affluent and better educated consumers not only are more likely to take full advantage of retailing innovation, but they also control more market infor- mation. Clearly the differential in market information is an important dimension in determining the "haves" and "have nots" in society. BCN technology can greatly facilitate the flow of market information, but not necessarily equally across all socio-economic groups. Katzman makes the point that new communication technology tends to increase the amount of information transmitted and received in a society and that all members of the society tend to receive more information that they would without it. However, the use of new communication technology will tend to be higher among those who already have higher 27 Rather than close economic and educational status. the gap between information-rich and infOrmation-poor consumers, the impact of BCN technology may well be to widen it. Those consumers who are presently more affluent and better educated will be more likely to take advantage of innovative shOpping services, thus 57 making them even more efficient Shoppers, and thus even more affluent. BCN technology may not serve to equalize the distribution of goods and services across society, but serve the opposite purpose. Because of the potential of BCN technology to reinforce the position of affluent consumer, it carries with it serious public policy implications. While the problem of unequal distribution of market information is most serious, it does not lend itself to empirical speculation. Unfortunately this analysis must be limited to less value laden descriptions of the shopping process. Bucklin did a survey of home- makers in Oakland, California to test the following general hypotheses about shopping behavior: 1. The consumer will shop more extensively where the cost of shopping is low. 2. The consumer will shop more extensively when she initially knows little about the product which she is buying and the stores that sell it. 3. The consumer will shop more extensively when the value of the product is high.28 In general Bucklin found support for all three hypotheses. The first hypothesis that consumers would shop more if the cost of shopping were lower is supported by evidence that consumers contacted more stores for a given product in a centralized downtown area or large plaza than in outlying, smaller retail centers. Similarly there was evidence supporting consumer's willingness to shop more for higher-priced items and when no brand was known or store preferred.29 58 Unfortunately Bucklin's study cannot provide quantitative estimates of the relationship between cost of shopping and propensity to shop. It is clear that if shOpping is made easier, or lower in cost, that more shopping will occur. The potential impact of BCN tech- nology should be primarily on this point, that is reduction of shopping cost. The fact that consumers shop more for higher-priced items and when they do not have strong brand or outlet preferences should remain the same whether or not a BCN system exists. This leaves two expected results of the application of BCN technology to marketing communication. First is a reduction in the amount of travel time associated with shopping, which is the communication substitute for transportation aspect of the application of BCN tech- nology discussed earlier. Second is an increase in shopping time resulting from the reduction of the cost of shopping as indicated by the Bucklin study. It is assumed that shopping via a BCN system is not more efficient than is Shopping in the traditional manner. Finding the desired goods in the store would be equiva- lent to waiting for the product to appear on the tele- vision screen. MarketingrResearch Marketing research provides the primary source of consumer feedback in the producer-consumer dialogue 59 discussed before. The most immediate applications of BCN technology to consumer marketing communication are research applications. A BCN system could totally revolutionize marketing research methodology, adding an entirely new method of communication in interactive television. Dordick and Hanneman describe five alternative marketing research strategies applying BCN technology. The first alternative is described as the combinational application which uses one cable channel in a one-way mode to provide program material to consumer homes. The material can be anything from product descriptions to commercials. Data can be collected using traditional techniques such as telephone or mail interviews. The second alternative is the dual channel coincidental method which allows for the Showing of commercials to a portion of the cable audience allowing some experi- mentation. The third alternative would allow character generated questions with a touchtone telephone reply. This alternative would partially solve the problem of the traditional data collection methods, but limits response to highly structured questions. The fourth alternative would be to use the distributed video disc to provide product and message preview information. Respondents would view the disc at their leisure and supply response data. The fifth, and most interesting 60 alternative is the full use of two-way cable technology. Two-way cable allows the combination of the best features of the other alternatives while allowing the response through the upstream digital return channel of the cable system.30 The use of two-way digital return interactive television allows the researcher to test concepts, test advertising copy, monitor television media behavior, and monitor purchase behavior. Probably the most serious limitation of the application of such technology to the marketing research process is the problem shared with mail research, that is who in the household is responding to the particular questions. This problem is a serious one for mail research. As Nuckols and Mayer point out, there is serious doubt of mail's ability to reach a particular houSehold member.31 The application of BCN technology to the consumer feedback process also allows the creation of consumer forums as mentioned before. BCN technology has the potential to remove some of the communication format control held by the marketer in the traditional research situation. A number of "hot lines" and other techniques have already been employed by some marketers in an attempt to open the channel of communication. Unfortu- nately, existing communication technology has limited such efforts to telephone and mail systems. 61 Because of the relatively low volume of research communiCation, there is little reason to expect that the application of BCN technology would have any direct effect on the life-styles of consumers. No doubt improved research techniques would provide better information to decision-makers, and as a result con- sumers would be provided with better products and ser- vices. Unfortunately this is an indirect impact, and extremely difficult to measure. Summary There are three potential changes in marketing communication related activity which have been identified that may be descriptive of the result of the application of BCN technology. The first of these is an increase in the amount of time spent watching television because of the increased attractiveness due to BCN capability. The increased attractiveness can be attributed to increased program diversity, but primarily to the addition of the return channel and the interactive mode. The second potential change is a decrease in the amount of time spent travelling in Shopping related activities. This is the communication-transportation tradeoff dimension of the application of BCN technology. The decrease in travel time is directly related to the number and type of shopping services available through a BCN system. 62 The third potential change is an increase in the time spent engaged in shopping activity. Because the cost of Shopping should be drastically reduced by the application of a BCN system, consumer interest in shopping should increase along with the time engaged in the activity. CHAPTER I I --NOTES lL. Michael Ray, "A Decision Sequence Analysis of Developments in Marketing Communication," Journal of Marketing 37 (January 1973): 29. 2Bent Stidsen and Thomas F. Schutte, "Marketing as a Communication System: The Marketing Concept Revisited," Journal of Marketing 36 (October 1972): 23. 3Ibid., p. 25. 4Leo Bogart, "As Media Change, How Will Adver- tising?" Journal of Advertising Research 13 (October 1973): 26. 5Ray, "A Decision Sequence," p. 38. 6Seymour Banks, "Trends Affecting the Implemen- tation of Advertising and Promotion," Journal of Marketing 37 (January 1973): 26—28. 7Karl W. Deutsch, "On Communication Models in the Social Sciences," Public Opinion Quarterly 16 (Fall 1952): 360. 8F. Craig Johnson and George R. Klare, "General Models of Communication Research: A Survey of the Developments of a Decade," Journal of Communication 11 (March 1961): 26. 9Ibido ’ pp. 17-240 10Alex Bavelas, "A Mathematical Model for Group Structures," Applied Anthropology 7 (Spring 1948): 16. 63 64 11Johnson and Klare, "General Models," pp. 15-16. 12Bogart, "As Media Change," p. 25. 13Anne W. Branscomb, "The Cable Fable: Will It Come True?" Journal of Communication 25 (Winter 1975): 48. 14Donald Edward Agostino, "A Comparison of Tele- vision Channel Use Between Cable Subscribers and Broad- cast Viewers in Selected Markets" (Ph.D. dissertation, Ohio University, 1974), pp. 128-29. 15John P. Robinson, "Television and Leisure Time: Yesterday, Today and (Maybe) Tomorrow," Public Opinion Qparterly 33 (Summer 1969): 222. 16Ibid., pp. 220-21. 17Edward M. Tauber, "Why Do People Shop?" Journal of Marketing 36 (October 1972): 49. 181bld., pp. 47-48. 19Ibid., p. 49. 20Peter L. Gillett, "A Profile of Urban In-Home Shoppers," Journal of Marketing 34 (July 1970): 40. 21Ibid., p. 45. 22W. Thomas Anderson, Jr., "Identifying the Convenience-Oriented Consumer," Journal of Marketing Research 8 (May 1971): 183. 23Fred D. Reynolds, "An Analysis of Catalog Buying Behavior," Journal of Marketing 38 (July 1974): 24Ibid. 25William H. Peters and Neil M. Ford, "A Profile of Urban In-Home ShOppers: The Other Half," Journal of Marketing 36 (January 1972): 64. 65 26Hans B. Thorelli, "Concentration of Information Power Among Consumers," Journal of Marketing Research 8 (November 1971): 432. 27Natan Katzman, "The Impact of Communications Technology: Promises and Prospects," Journal of Com- munication 24 (Autumn 1974): 49-57. 28Louis P. Bucklin, "Testing Propensities to Shop," Journal of Marketing 30 (January 1966): 22. 29Ibid., p. 27. 30Herbert S. Dordick and Gerhard J. Hanneman, "Utilization of Broadband Communications Network Tech- nology for PrOprietary Market Research," University of Southern California, Los Angeles, California, 1975, pp. 17-19. (Mimeographed.) 31Robert C. Nuckols and Charles S. Mayer, "Can Independent ReSponseS Be Obtained from Various Members in a Mail Panel Household?" Journal of Marketing Research 7 (February 1970): 94. CHAPTER III TECHNOLOGICAL FORECASTING METHODOLOGY Forecasting New Technology Technological forecasting is a relatively recent development among the tools of scientists and social planners. Technological forecasting as a serious pro- fessional activity has been developed primarily by those engaged in the administration of military research and development since 1965. Bright and Schoeman describe the state of the art of technological forecasting prior to 1965 as follows: There was virtually no published methodology for forecasting technology. Past practice had been to use expert opinion. Opinion was not a very satisfying predictive device, as was evidenced by the highly uneven record of scientists and engineers with unassailable records of technical competence. a .After 1967, technological forecasting became a popular topic for seminars and short courses along with related topics such as research and development planning, and corporate planning. In 1969 three journals involved with technological forecasting were introduced, Futures, Technological Forecasting and Social Change, and Long Range Planning. 66 67 The major reason for the general neglect of technology among economists and social scientists, as Ayres points out, is the widespread notion that technologi— cal change is inherently unpredictable. Largely responsible for this, according to Ayres, is the popular mystique of science as having "an incompre- hensible internal dynamism relying mainly on the work- ings of creative genius--resulting in perpetually unexpected breakthroughs."2 Recently, however, science and technology are becoming less regarded as a phenomena themselves, and more as a vital feature of society and economic change. Martino provides a working definition of techno- logical forecasting as "prediction of the future char- acteristics of useful machines, procedures or techniques."3 Ayres adds to the definition of a forecast as being "a reasonably definite statement about the future, usually qualified in the sense of being contingent on an unchang- ing or very slowly changing environment."4 Ayres also makes the point consistent with Reid's view of tele- communication research, that technology is generally created in response to societal needs.5 Thus, it would seem certain that technology is not driven by a mystical technological dynamic, but rather a societal demand for technological progress. Heiss, Knorr, and Morgenstern describe four kinds of technological advances that should be recognized in .’ 68 any attempt at predicting the future impact of tech- nology. First are marginal improvements in known technologies. Examples of such marginal improvement are in the increased mileage and durability of automobile tires because of belted tire constructions and cord material, and in the availability of television program material because of the development of the UHF frequency spectrum. Second are the combined application of known and improvable technologies in order to perform a com- pletely new task. The example given by Heiss et a1. is the NASA plan to put a man on the moon. Predictions about these types of technological improvement are not particularly difficult, although the exact timing is always subject to question because the level of financial and intellectual effort required are not always readily available. The third kind of technological prediction is much more difficult, that is technology developed because of the great benefits it promises. This is the case Where a large amount of resources are devoted to the develoPment of technology in order to achieve certain economic, military, or social goals. Generally new te<3hnok>gy must be invented to solve the problem, such as a cure for cancer, making it extremely difficult to Predict . 69 The fourth kind of new technology is that which nothing is currently known about and comes as a complete surprise. Obviously this kind of technological change is impossible to predict. Although, as Heiss et a1. point out, there have been relatively few complete sur- prises in new technology. Among those mentioned are aspirin in 1900, the nuclear bomb in 1930, and lasers in 1950.6 Technological forecasting methodology is then best suited to predict the first two kinds of techno- logical change, marginal improvements in known technolo- gies and the combined application 'of known and improvable technology to perform new tasks. BCN technology is clearly in the second category, that is combined appli- cation of known technology to perform new tasks. AS described in Chapter I, BCN technology represents a Combination of cable television technology, computer and information system technology, and communication Satellite technology. BCN technology represents far More than just the marginal improvement of existing television technology. BCN technology also makes POSSj-ble completely new applications of telecommuni- cations. Technological forecasting methodology can be conveniently divided into four methodological groups, inclUding dialectical methods, teleological methods, 70 experimental and empirical methods, and analytical methods. Each of the methodological categories will be considered separately. Dialectical Methods The dialectical approach is founded on the notion that both the future and past history are the result of a sequence of conflicts. The dialectical method itself is divided into two major approaches, the scenario approach and the Delphi Method. The scenario approach is perhaps the most commonly used of all of the technological forecasting methodology, yet it iS by far the most mystical. The scenario approach is not capable of systematic explanation as are the other methodologies described here. The approach involves the consultation of a large volume of relevant material and literature, and then the application Of intuitive judgment to describe the future. The Scenario approach is much the same as that of the his- tOrian. Judgment and intuition are crucial in develop- ing the forecast. It is very easy to dismiss scenario forecasts, granting them only some mild entertainment value. How- ever, scenarios form the basis for technological fore- cast, 01‘ any statement at all about the future. As Heiss et al. point out, the scenario provides the frame- work for speculation and is necessary to begin the 71 process of making any predictions.7 Although, it is probably safe to say that scenarios describing short-term predictions about relatively slow changing phenomena will in general be more credible than longer-term predictions. The scenario itself is an exercise in the sequential plotting of events as they could happen. Generally when the scenario approach is used several different future conditions are described, with one representing a medium projection, and others representing a high and low projection based upon various associated phenomena. The Scenario itself is an intuitive extra- polation of the emerging trends evident to an expert. A more formal dialectical method is the Delphi Method. The Delphi Method is probably the best known Of technological forecasting methodologies. The Delphi Method has been used to make forecasts of future con- ditions in a variety of industrial and governmental Settings, including the assessment of possible levels Of demand for a company's product. Jolson and Rossow found the Delphi Method to be preferred over the hunch Of a single decision maker or a consensus of a group f<311<>W1ng face-to-face discussion in assigning prior PrObabi-lities in a marketing decision to be made under uncertainty , 8 The Delphi Method, developed by the Rand Cor- Poration, is a unique method of eliciting and refining group judgment. The salient features of the method are _: _- “1.... an. _ ._ H. _.~.. _ 5:... 72 anonymity, controlled feedback, and group response. During a Delphi sequence, the group members are not made known to each other. The interaction of the group members are handled completely anonymously. The group interaction is conducted through the use of question- naires, with the individual or agency conducting the sequence extracting the relevant pieces of information and presenting back to the group. This is the way that feedback to the group members is controlled. Instead of reporting a simple majority viewpoint, the Delphi sequence presents a statistical reSponse ensuring that the opinion of every panel member is taken into con- sideration. The basic theory behind the Delphi Method is that with repeated measurement the range of responses Will decrease and converge toward the midrange of the distribution, and that the total group response will successively move toward the correct or true answer. It should also be mentioned that the Delphi Method is intended only for use with groups of experts in the appropriate field. Convergence with nonexpert panels Simply does not occur. Martino outlines a number of advantages and dis- advantages associated with the Delphi Method. Some of the advantages include: (1) the sum of the information av"inable to a group is at least as great as the infor- mation EVailable to its members, and (2) the number 73 of factors which relate to a given area which can be considered by a group is at least as great as the number available to any one group member. While it is true that a group does involve the pooling of resources, it has been found experimentally to be true that a group tends to be more willing to take risks than an individual. This last advantage is especially important in the forecasting field. Among the disadvantages outlined by Martino for the Delphi Method is the opposite of the pooled information advantage, that is, there is at least as much misinformation available to the group as to any Of its members. Other problems include the potential for exerting social pressure on its members, which can contribute to the "bandwagon" effect. A group can become more interested in reaching agreement than in developing a well thought out forecast. It has also been shown experimentally that it is not the validity Of arguments in a group that carry the group opinion, but the number of arguments. This phenomenon makes it possible to influence the group simply by sheer Volume of argument, and opens the possibility of one indiVidual dominating the group, even though he is anonymous. It is also possible that certain group members have vested interests in certain points of View and spend their time attempting to convert the 74 others, rather than to reach a well thought out forecast. It is also true that a group of experts that come from similar backgrounds are likely to share a common bias, especially in the area of technology.9 While there are numerous variations and examples of the Delphi Method which could be considered here, they are beyond the scope of this brief introduction to technological forecasting methodology. Teleological methods are considered next. Teleological Methods Teleological methods explicitly recognize the interaction between the forecasts made and the future itself. Teleological methods are considered by Ayres 10 as "inventing the future." Teleological forecasts are goal oriented, in that they typically assess future goals , and then work back to the present. Heiss et al. Suggest that a better term for this kind of technological forecasting is technology assessment.11 The essential Point of such forecasting methods is that the world without a particular forecast is different from the World with the forecast. The future is directed toward the Particular goal. Technology assessment involves two major tech- niques, the normative "backcast" such as PERT analysis, and cost-benefit analysis. PERT analysis, or Program Evaluation and Review Technique, is an advanced approach 75 to planning complex projects involving many interrelated and interdependent tasks. In PERT, a sequential network of activities is constructed with time estimates for completion of the necessary activities. The PERT net- work is normally represented by a flow diagram consisting of the activities and interdependencies. Activities are nortnally circles, and the relationship between events are normally lines. The "critical path" becomes the longest path in time to complete the project. Obviously the use of a technique such as PERT in technological forecasting is goal oriented. Another teleological method is the cost-benefit analysis. Such analysis applied to technological fore- casting focuses on a measure applicable wherever a functional goal can be defined referred to as "cost- effectiveness." Cost-effectiveness refers to a technique for demonstrating the tradeoffs between sunk costs and results achieved. The application of cost-effectiveness as a measure normally implies that a choice must be made from a number of alternatives. Ayres suggests a number of criteria which can be applied to optimize cost- effectiveness: 1- To maximize "military value" within a fixed budget 2- To maximize the probability of achieving a designated capability within a fixed budget 3- To minimize the cost of achieving a designated capability _-‘ 76 4. To maximize technological cross-support and spinoff while achieving a minimum goal within a fixed budget 5. To do any of the above subject to additional constraints. Techniques such as cost-effectiveness and PERT come directly from military planning. Experimental and Empirical Methods Experimental and empirical methods are data orzieented technological forecasting methods. While a rnnmmoer of quantitative analyses developed in statistics, econometrics and survey research can provide data on whimzll to base forecasts, the technique most commonly assc>c:iated with this method is the time series analysis. Time series analysis is a technique which allows past. cexperience to guide future expectation through the numerical representation of some set of empirical obsezrxvations. The fundamental process involves the extraalpolation of trends from a mathematically described Series of historical points. The technique is often objected to, as Martino points out, because the fore- caster cannot describe events which lie beyond the range <>f his (data points. Martino argues that such extrapo- lation is possible because of continuity. For example, if in some area of technology there has been a con- 1:inuous jprogression of successive technical approaches, it is not unreasonable to expect it to continue. To argue Otherwise is to advocate discontinuity, or 77 that despite a more or less regular innovation in the past, the present represents a halt.13 There exists a vast amount of literature describ- ing time series analysis, with the majority of the recent literature highly mathematical in nature. The reason for this, according to Heiss et al., is because of the extraordinary complexities which a time series may show.14 At the risk of some oversimplification, Ayres describes two simple types of curves which are used either singly or in complex combinations. The first is the periodic or cyclic wave form with a single dominant period, such as a sine wave. The second are cOntinuous curves with at least two derivatives, and at most one minimum or maximum and no points of inflection. Examples of this include a simple straight line or an exponential curve.15 Simple linear regression is an example of the application of the latter type of curve. Common mathematical smooth- ing techniques such as moving averages and exponential smoothing can also be added. It is beyond the scape of this brief introduction to consider these techniques in any detail. It should also be mentioned that despite the apparent quantitative sophistication of some of the curve fitting procedures, forecasts based on the curve are still dependent upon a number of subjective factors. 78 Subjectivity is involved in both the selection and pro- cessing of the input data and the interpretation of the output. Heiss et a1. mention two potential pitfalls in applying time series analysis in technological forecast- ing outside the ability of the mathematical function to adequately describe the trend. First is in an aggre- gation series there is usually a great change in compo- sition. An example is in the Gross National Product which in 1935 did not include a number of very large industries today, such as electronics, television, computers, jet engines and nuclear reactors. Second is in an individual series there may be a problem with homogeneity through time. An example of this can be seen in the substitution of plastics for metals in a large number of products.16 Time series analysis is a valuable tool for (determining the limits or boundaries around a particular forecast. However, time series analysis Should be applied with caution . Analytical Methods Analytical methods involve the model oriented approach to technological forecasting. The analytical method is termed by Martino as the "black box" approach to: forecasting.17 Analytical forecasting models are generally theoretically based models trying to explain 79 the interrelationships between variables. The variables can be endogenous or explained within the model, as well as exogenous as inputs that are historically given or inputs under the control of the investigator. The distinction between empirical methods and analytical methods is often unclear, but is generally defined in terms of the starting point of the forecaster, whether it be in data or a subjective view of some relationships that can be converted into a model. Generally speaking an empirical method must involve some kind of model, and an analytical method must ultimately involve some data. To employ the analytical method it is assumed that it is possible to identify the elements of the par- ticular process and all interactions. Analytical models can only provide forecasts when certain elements in the situation are going to change. Analytical methods can be divided into two general types, deterministic and stochastic. Deterministic models are generally mathematically SOphisticated models which offer unique solutions to a forecasting problem when the situation «can.be sufficiently simplified to fit the requirements «of the model. Stochastic models, while mathematically sophisticated, do not offer unique solutions, but rather offer specific case results. Stochastic models make use of various random processes and are normally considered in terms of computer simulation. Computer 80 simulation is capable of handling far more complex situations than is a deterministic mathematical model. As Martino indicates, the current state of the art of technological forecasting analytical models is quite primitive. Existing models have only limited utility for very practical forecasting purposes.18 A relatively well-known example is the Limits to Growth computer simulation model developed by Meadows, Meadows, Randers, and Behrens.19 The main reason for the lack of analytical models is the amount of effort necessary to develop them. But as Martino speculates: The primitive state of current models, far from being a cause for discouragement, should actually be a cause for hope. If the limited amount of effort which has been expended can produce results even this good, we can expect that much better results will be obtained when considerably more effort is applied to larger quantities of data in many varied situations. Before continuing the discussion of computer simulation applied to technological forecasting, several published examples of forecasts of telecommunication technology will be presented. Dialectical Methods Applied to Telecommunication The "expert opinion" approach is by far the most popular technique for forecasting the future of tele- ¢3cummunication technology in general. Numerous publications describe long lists of future applications and uses for 81 telecommunication technology similar to the lists described in Chapter I. Hinshaw provides a provocative pair of scenarios describing the future as a result of the impact of tele- communication technology. In the first scenario as social disorganization and environmental degradation reach new highs, a new and powerful union of social and behavioral science and the technological sciences emerge. After all homes are required to have at least one basic two-way terminal, social stability is brought about by strictly imposed government regulation. In the second scenario Hinshaw describes a far more attrac- tive future where telecommunication technology is recog- nized as a medium for the creation of wholly new com- munities and as a tool for exchanging information. Instead of less human interaction as in the first scenario, more interaction develops.21 While Hinshaw expresses a preference for the latter Scenario, he loelieves the first to be the more likely future. He provides an extremely interesting bit of speculation and a very good example of the scenario approach. Baran, as described in Chapter I, provides one (of the first examples of expert Opinion forecasting telecommunication technology as a marketing setting. Baran describes the future as follows: Imagine consumers in the year 2000. Much of the shopping will be done from the home via TV display. Think of the screen as a general 82 purpose genie. Pressing a few buttons on a key- board allows interaction with a powerful infor- mation processing network. The information network sends back a modified image to the TV display in response to selections. . . . 22 Carne in less interesting language outlines the capa- bilities of the new communication technolOgy by promising tomorrow's businessman the ability to: 1. Talk face-to-face with associates regardless of distance. 2. Participate in nationwide meetings without leaving his office building. 3. Perform a large number of executive functions from his home. 4. Have access to unlimited computer resources. 5. Own and operate private telecommunication facilities. These capabilities are predicted by the mid-1980's. Barnett and Greenberg argue for the justification of the cost of $60 per home for wiring the city with the following points. First is an increase in the number of available channels. Second is an increase in the number and diversity of program offerings at a reduced cost. Third is improved picture quality. Fourth is a cost saving for both homes and broadcasters because it is less expensive to wire a city than it is to build a ‘transmitter and purchase antennas. Fifth is the benefit of frequency spectrum; saving, and finally increased flexibility in communication opening the door for additional services such as Shopping. They do point curt some negative points, however, in the increased cost to build in rural areas, the fact that it is cheaper 83 use conventional broadcasting systems in the larger an areas, and high programming costs which would be urred to fill the twenty channels.24 Despite all of problems, including the regulatory environment, nett and Greenberg intuitively conclude that "the mise and potentials of wired city television and er wired city services are so extraordinary that imately we may have them."25 Eppirical and Analytical Methods Applied to Teiecommunication Young's economic analysis of the telecommunication ustry is an example of a mixed empirical-analytic roach to forecasting. The analysis concentrating on ‘expenditure on research and development in the ustry found expenditure in the category to be very h, yet returns to capital and labor in direct pro- tion to be very low. It should be mentioned that ng's analysis restricts the telecommunication ustry to the telephone industry and employs econo- ric regression analysis.26 Ohls attempted to develop an investment theory cific to the CATV industry. Ohls, however, did not e use of any empirical data, restricting the consid- tion to economic theory only. It would be necessary add data to apply the Ohls model to a forecast of potential for economic growth for CATV, assuming 84 rational investors. Ohls considers a number of factors in his optimum investment analysis including the number of channels and the amount of local-origination program- ming, the number of common carrier uses such as leasing channels, and the acceptance of advertising.27 Perhaps the best example of technological fore- casting applied to any BCN related technology is the Park "impact model." Park began with the assumption that the future impact of cable television on broad- casting depends in large part on how cable continues to grow. The model begins by fitting a logistic growth curve to the penetration of cable television. A logistic growth curve is generally a good descriptor of growth within a finite limit and is given by the following equation: Y = ea-b/T The size of the growing entity is given by Y, here the number of cable television subscribers. T is the elapsed time since the growth began, and a and b are parameters of the logistic distribution. The base of natural logarithms is standardly noted as e. The ultimate, or maximum, number of subscribers is a function of the a parameter and is expressed as ea. This obviously will depend on other factors such as the number of households in a given service area, H, 85 and the fraction of all households expected to subscribe, Fi' if the system offers services of type i. The ulti- mate size of the system, then, can be expressed by the following equation: a = u e FiHe An error term, u, is introduced to represent all other factors. Expanding the b parameter to represent a possible curvilinear relation across the number of households, the following equation is derived: _ 2 b — b1 + sz + b3H The b parameter represents a how "stretched out" is the growth curve. As Park points out, the larger the number of households the more stretched out the growth curve is expected to be. Combining these three equations Park derives the overall estimation equation used in the analysis. The equation is expressed as follows: log Y = log Fi + log H + bl(-l/T) + b2(-H/T) + b3(-H2/T) + u The equation becomes an appropriate subject for ordinary least square regression.28 Using data for a large number of cable systems obtained from such sources as the Television Factbook, 86 Park is able to estimate the relative impact of a number of variables including the number of distant signals carried by a cable system, the attractiveness of television programming as indicated by audience shares, and other variables such as public service pro- gramming. Upon completion of the analysis, Park reached four major conclusions: 1. Reduction in aggregate station revenue due to cable penetration is perhaps not large enough to justify any great concern. . . . 2. Stations in larger markets, now sheltered by FCC policy, would on average be little hurt by unrestricted cable growth. 3. Stations in smaller markets, for which FCC policy now provides no protection, would suffer severe revenue restriction due to cable at ultimate penetration. . . 4. In the near term, say through the 1970's, non- network UHF stations stand to gain substantially from cable growth, because cable puts them on the same technical footing as competing VHF stations.2 The Park Impact Model represents the most sophisticated forecasting effort of BCN technology yet published. Computer Simulation The availability of computer and information processing resources to those engaged in technological forecasting makes the application of computer simulation seem like a logical extension of the resource. Computer simulation seems at an intuitive level at least to be superior to the rather simplistic "seat-of-the-pants" approach evidenced in some dialectical methods. The direct application of deterministic mathematical models 87 en yields unsatisfactory results, largely because of difficulty in adequately describing the real world .uation in mathematical terms. Computer simulation 'ers to some degree a solution to this problem. Abelson says that the advent of simulation tech- ues to the social sciences has been so widely heralded .t "many practioners and students of a variety of dis- .lines are clamoring to use these techniques."30 .zenbaum, in an even stronger statement, says that .e time is not far off when no social science theory .t cannot be reduced to a computer model will be given ' credence or respect because its non-reducibility 31‘ Certainly what would I only mean its ambiguity." true of social science theory ought to be partially 1e for technological forecasting. The general approach adopted in computer simu- ;ion techniques is often referred to as functional (lysis. Kline, in describing the general leitmotif communication research, uses a passage borrowed from 'ton to describe functional analysis: "Functional Llysis, to a great extent, is concerned with examining »se consequences of social phenomena which affect the fmal operation, adaptation, or adjustment of a given :ial system: individuals, subgroups, social and .tural systems."32 From the point of view of Imunication research, the functional approach 88 clearly adopts the point of view of the strategist, and is no doubt the orientation that the technological fore- caster would be most comfortable with. This conceptual framework makes the computer simulation approach very attractive to the applications-oriented model developers because it forces problems into their "real-life" con- texts. Schultz and Sullivan describe the use of computer simulation as a research methodology in three ways. First, computer simulation can be a technique for theory building and comparison which accounts for the interest in the technique among scientists and scholars. The second major use is as a teaching and training device such as the popular marketing games. Third and last is the use of computer simulation as an aid to practical decision-making in organizations which is the business 33 It is and marketing application of the technique. this last use of computer simulation that makes the technique appropriate to technological forecasting. Computer simulation is not a panacea, or magical machine that solves all problems by merely "feeding them into the computer." Computer simulation is nothing more than a pggi, albeit a unique and sophisticated kind of tool. Implicit in the definition of computer simulation is the concept of isomorphism between some real world phenomena and a computerized model. Schultz and 89 Sullivan describe a computer simulation as an operating 34 Abelson defines computer model of a real system. simulation as the "construction of models of behavior which can be exercised or run on a computer so that certain aspects of the computer's symbolic behavior imitate or simulate real-world behavior."35 Pool describes phenomena which can be simulated as any phenomena which has a describable pattern or nonrandom behavior. For a simulation to be computerizable, "there must exist somewhere a complex structure of propositions and/or data values. But note that we have had to use the expression and/or. It is not necessarily true that a Simulation is no more useful than its data."36 The minimal conditions for a computer simulation of any human behavior process are set forth by Dutton and Briggs. Among these are: 1. Examines a behavior process. 2. Gives a theory which describes and explains the process without amiguity. 3. Shows how the process is affected by its environment. - 4. Be formulated in such a way that inferences about the process may be verified by obser- vations. In order to be computerizable, then, it is necessary to Ihaye an empirically verifiable and unambiguous system. In order to be simulated, it is necessary to have a «describable process and an "open system." The idea (of the system being affected by its environment is 90 central to the conceptualization of computer simulation. A closed system would be nothing more than a continuous processing loop. Abelson specifies two basic requirements for models or processes that would lend themselves to com— puter simulation. First is the requirement for a dynamic model, that is something that will happen over time. Second is the requirement for well-specified independent and dependent variables, as well as a clear statement of the relationships between these variables.38 Abelson defines six features possessed by a well-specified dynamic model that is capable of being simulated. These features are: 1. Units: a set of elements or entities defining the units of concern to the investigator, the units which embody or "carry" the processes of interest. 2. Properties: a set of variables (and constants) which attach to each unit and define its state at a given moment. 3. In uts: a set of stimuli or task variables putt1ng the postulated processes into relevant motion and possibly intervening during the later course of these processes. 4. Processes: a set of specifications of what is supposed to "happen," that is, of which prOper- ties of which units are to change as what function of the inputs of other units. 5. Phasin : the organization of the sequence in which the processes are to occur. 6. Consequences: one or more of the final proper- Eies of key units or aggregations of units: the crucial events, outcomes or states of affairs about which the model is capable of making predictions. Computer simulation makes possible a more realistic approach to problem solving. Newell and 91 Meier describe heuristic problem solving in terms of "simulating" the techniques and procedures that might be employed by intelligent problem solvers.40 Computer simulation can lead to the generation of useful relationships that may not otherwise be apparent as well as incorporate the problem solving technique of the intelligent investigator or forecaster. Computer simulation, then, can serve a heuristic function on at least two levels, first by actually generating new relationships and guiding further research, and second by incorporating the process into the model itself. Pool, Abelson, and Popkin suggest that computer simulation can provide a way of analyzing fanciful questions scientifically, that is the "what if" question. Computer simulation allows the investigator to model a hypothetical world, with all of the relevant complexities, and manipulate the model for any one of several purposes; "such as prediction, postdiction for analysis, exploration of alternatives for sensitivity testing, and exploration of policy alternatives."41 The ability to ask the "what if" question is one of the major attractions of computer simulation to techno- logical forecasting. Schultz and Sullivan have succinctly expressed the advantages of computer simulation: 92 1. It provides a method for vicarious experimentation. 2. It allows more complete models with greater degree of complexity--a holistic approach, and can be eclectic, permitting combinations of theories and methodologies. 3. It permits manipulation of an iconic model of time. 4. It allows random or pseudorandom variables to be introduced directly into the model. The disadvantages of computer simulation are: 1. It provides specific case results requiring replications to produce more general results. 2. It usually requires more effort in constructing the model. 3. It can lead to more apparent realism and con- sequently greater danger of forgetting the limi- tations of the model. 2 The applications of computer simulation are gen- erally divided along two dimensiOns: (1) application area, that is social science versus business appli- cations, and (2) level of analysis, that is macro- behavioral processes versus microbehavioral processes. Some examples of microbehavioral processes, where the individual is the unit of analysis, include socialization processes, other reference group processes, individual processing of social communication, and social exchange considerations in decision-making. Some examples of macrobehavioral processes, where the group is the unit of analysis, include social organization, population dynamics, polity processes, and information dissemi- nation. The TIMMOD simulation which is described in Chapter V is a microbehavioral Simulation, that is the individual is the unit of analysis. 93 When computer Simulation is considered in its business application, that is an aid to practical decision-making, Schultz and Sullivan suggest four general ways that computer simulation can help. Among these are: 1. Can increase the general understanding of interest as an aid in making future decisions. 2. Can guide the development of alternative plans or courses of action. 3. Can evaluate the merits of alternative courses of action. 4. Can provide for verification and control, that is can provide norms and/or standards against which performance can be measured. All four of these suggested aids can assist in the techno- logical forecasting situation. Newell and Meier suggest several applications for computer simulation in business. Among these are inventory systems, job shop scheduling, PERT networks, risk analysis and capital investments, waiting lines, and forecasting.44 Kotler and Schultz describe four basic uses for simulation in marketing. These include: 1. Simulation as behavioral modeling. 2. Simulation as a way of introducing and handling uncertainty. 3. Simulation as a computational technique for measuring parametric sensitivity. 4. Simulation as a heuristic technique for finding an approximately optimal solution. These uses for simulation, of course, are very similar to the advantages of computer simulation described by Schultz and Sullivan. Kotler and Schultz mention one of the primary disadvantages of simulation techniques 94 when they say, "Although simulation methodology often produces new and useful perspectives, the analyst should always try to anticipate whether the view will be worth the climb."46 Abelson nicely summarizes the discussion of com- puter simulation when he says that the power and novelty of the computer simulation technique lies in its ability to synthesize.47 It is this ability to synthesize that makes computer simulation attractive to technological forecasting. Before considering some examples of computer simulation in areas relating to telecommunication, the problems of designing computer simulation experiments and the very difficult problem of computer simulation model validation will be discussed. Problems in the Design of Computer Simulation Experiments Naylor, Balintfy, Burdick, and Chu outline nine major steps necessary in planning a computer simulation experiment. The first of these is a formulation of the ‘problem. Research objectives must be formulated which include questions to be answered, or hypotheses to be tested, or effects to be estimated. Almost as impor- tant as a good problem statement is a decision on the objectives of the research, and the criteria for evalu- ating the degree to which the objectives are fulfilled 95 by the experiment. In a technological forecasting situation there may not be hypotheses in the same sense as in the development of theory, but a clear statement of either the questions to be answered or the effect to be estimated is mandatory. The second step is the collection and processing of real world data. It is at this step where the dis- tinction between analytical and empirical methods blur. The use of data, or information, at this early stage is vi tal in understanding the process to be simulated, and may well suggest hypotheses and mathematical models to be employed later. Data for a computer simulation must be collected and recorded. It is'also important that aata be converted to machine sensible form for later input into the model. The third step involves the formulation of ma thematical models. This step consists of three parts, including the specification of components, the specifi- eatiion of variables and parameters, and the specification of functional relationships. Among the considerations in this step are a determination of the number of variables to be included in the model, the mathematical coInplexity of the model, the computational efficiency of mathematical procedures, the realism, and the com- Datibility of the model with the type of experiments to be conducted with it. 96 The fourth step involves the estimation of ¢3g>erating characteristics from real world data. It is eat: this point that the rules for the manipulation of greeal world data are determined for the model, for eeacammfle the use of theoretical probability distributions to represent certain empirical phenomena. The fifth step is the evaluation of the model aarraci parameter estimates. This step is considered by Israaagflor et al. as the first step in testing the model. S ix important questions need to be asked at this point: 1. Have we included any variables which are not pertinent in the sense that they contribute little to our ability to predict the behavior of the endogenous variables of our system? Have we failed to include one or more endogenous variables that are likely to affect the behavior of the endogenous variables in the system? Have we accurately formulated one or more of the functional relationships between our system's endogenous and exogenous variables? Have the estimates of the parameters of the system's operating characteristics been esti- mated properly? Are the estimates of the parameters in our model statistically significant? On the basis of hand calculations how do the theoretical values of the endogenous variables of our system compare with historical or actual values of the endogenous variables? The sixth step is the actual formulation of a coInputer program. This begins with a system design nor- I“Willyrepresented on a flow chart. The program is then QOclea and debugged . The seventh step is one of the most difficult, validation. Generally Speaking two tests are normally 97 considered appropriate; how well the model postdicts historical data if it is available, and how well it Ioaredicts future data. The problem of validation will I>ee discussed in more detail later. The eighth step involves the design of simulation experiments. As with other forms of experimentation, the problems of determining an appropriate experimental design are the same for a computer simulation experiment. Treatment and control conditions must be allocated so that the research hypotheses can be adequately tested. The ninth and final step involves the analysis 0 f the simulated data. This is the step where the com- puter simulation is run and the simulated data are generated. The appropriate test statistics are then Calculated as required by the experimental design. The results of the simulation experiment are then i 1'1 terpreted . 49 Naylor discusses four basic design problems in <2anputer simulation experiments. The first of these is file problem of stochastic convergence. Since most 8ilmulations are designed to produce estimates of popu- lation averages, it is reasonable to expect random fluctuation in the averages generated. Remembering tlhat simulated characteristics are normally generated through a stochastic process, "sampling" error is eXpected. The problem is that convergence toward the 98 "true" average occurs slowly. The mathematical relation- ship is an inverse square as shown by the following equation: In this equation it would require four times the n, or sample size to reduce the standard error of the mean, S by one half. The standard deviation of the popu- e’ lation is represented by S. The second problem in simulation experiments is the problem of size or too many factors. This problem is common to all experiments. It is very easy to design an experiment with too many factors and levels to handle reasonably. It is easy to let the number of cells required in an experiment to get very large. The problem is probably not as critical in real world experiments because the limitations imposed by available sample sizes. It is easy to overlook the problem when working with a computer. The third problem is that of motive. It is important that the experimenter design the experiment so that it best satisfies the objectives. Generally there are two types of experimental objectives that can be identified; first is to find the combination of factor levels which minimize or maximize the response variable in order to optimize some process, and second 99 is to make a general investigation of the relationship of the response to the factors in order to determine the underlying mechanisms governing the process. The fourth problem is the problem of multiple response. This occurs when there are too many response variables. One way of handling the problem is to run an experiment for each response variable. If multiple responses are required for the experiment, then the problem becomes severe because there is very little methodology developed to handle the problem.50 The difficult problem of validity for a computer simulation model is much the same as for any other kind of model. The fact that computer simulation often appears somewhat mysterious because it is so complex seems to draw unusual attention to the problem. The problem of validating any model involves a host of practical, theoretical, statistical, and philOSOphical questions, which are well beyond the consideration here. The most satisfactory approach to computer simu- lation model validation is the multistage validation procedure proposed by Naylor. The first stage calls for the formulation of a set of postulates or hypotheses describing the behavior of the system. These postulates are based on the researcher's "general knowledge." The second stage involves an attempt to verify these postulates subject to the limitations of existing texts. ..' ‘ W“ ' n ".5771 100 In the case of postulates which cannot be verified empirically, they must either be labelled tentative or declared false and abandoned. The third stage consists of testing the model's ability to predict the behavior of the system. As mentioned before there are two general alternatives, historical verification and verification by forecasting.Sl Numerous statistical tests are available to determine how well a computer simulation model postdicts or predicts, including chi- square goodness-of-fit, analysis of variance, non- parametric tests, and correlation and regression. Mass Media Related Computer Simulation Models With perhaps the exception of the Limits of Growth Model described earlier, there are virtually no technological forecasting computer simulation models in the published literature. There are, however, a number of mass media related computer simulation models pub- lished in the literature which are of interest. The grandfather of all the mass media simulations in the Simulmatics Media Selection Model developed in 1962. The Simulmatics Model was designed as a tool to help solve the advertising media allocation problem. In fact the most sophisticated mass media simulations are designed for this purpose. The basic purpose of the Simulmatics Model was not to find the optimal plan, but to test one or more 101 proposed plans. The basic input to the model is a sample of 2,944 hypothetical persons who supposedly represent a cross-section of the American population. Individual media choices are determined probabilistically from socio-economic characteristics. The process con- sists of drawing a random two-digit number and cycling it through the various media. If this number is greater than the estimated probability of watching the particular media, then the individual is recorded as having not seen the media or show, and if it is less, he has been exposed. The process is continued until all of the hypothetical population is exhausted and the appropriate number of time periods have been considered. The output is an estimate of the audience profile and the reach and frequency characteristics of the prOposed media schedule.52 The COMCOM (COMmunist COMmunication) model is a simulation of a mass media system in the Soviet Union described by Popkin. Ideally this kind of model would input demographic, attitudinal, and media data from a panel survey as is done using the Simulmatics Model in the United States. However, this kind of information simply is not available, so the model must estimate ‘these parameters. The model goes through two passes, first to develop a hypothetical description of indi- 'viduals who constitute a sample of the relevant popu- lation, and.second a series of messages which constitute ‘ Awr'.$‘.4.‘ _ 911 rt)“ ‘1 a 5 I 102 .w. The basic exposure measure is derived from a .te Carlo technique. The output of the model is an .imate of the cumulative audience over n periods of le for the particular communist country.53 Kramer developed a comprehensive simulation of : public information campaign in Cincinnati, Ohio to Lcate the pOpulation about the United Nations in 1947 l 1948. Survey research data indicated only a slight :nge in attitudes as a result of the campaign toward 2 United Nations, but substantial changes in attitudes .ut such things as war, relations with the Soviet Union, 1 control of the atomic bomb. The simulation was signed to predict reach and frequency of message flow >m known facts about the population, media habits, and a placement of messages in the mass media. The simu- :ion is divided into two stages. The first stage, >wn in Figure l, is the data disaggregation and inte- ltiOh stage where such diverse data as population data, lience cumulation and duplication data are combined describe the model population. The model begins with sic population data such as age, sex, geographic [ion, and literacy. Parameters are then estimated to L1 in the cells where data are missing. Audience auts for each media vehicle are provided and an average >bability exposure is calculated. The actual exposure stribution is a two-parameter beta distribution. To 103 Population Estillte In t ‘ Pare-eter- pu Where Needed 1. Store Vhluee in? i l, — g E f! Audience Calculate Inputs for Exposure 2 each Vehicle Probability ! i L Store Values in P Determine Beta Parameters Audience Calculate Inputs for ‘ Accuuletion Vehicle Pairs Probability anmxumnmnmn A‘J Fig. l. Kramer Mass Media Simulation Flow Diagram Stage One: Data Disaggregation and Integration SOURCE: John F. Kramer, "A Computer Simulation of .Audience Exposure in a Mass Media System: The United Nations Information Campaign in Cincinnati 1947-48" (PhJD. dissertation, M.I.T., 1969). 104 allow for audience accumulation and duplication the pairwise audience duplication data are entered into the model and probabilities are assigned to population mem- bers. The population is then ready to be passed to the second stage, which is shown in Figure 2. The second stage is the processing messages and reporting exposures stage which begins by converting calculating message exposure probabilities from a message audience distri- bution. The message audience distribution is determined from message audience input for each theme to be con- sidered in the simulation run. The message themes are then entered into the model including the timing. This, in combination with the model population, is used to determine the product of message exposure probability, the vehicle exposure probability, and a "format" factor, to get the net probability of exposure. The net exposure is then aggregated and the model is cycled through the next time period. The themes for the model were determined from a content analysis of the messages running during the cam- paign. Twelve themes were identified, such as U.S./ U.S.S.R. hostility, and described in terms of attitudes and exposure across demographic subgroups of the popu- lation. The model itself can handle up to 64 media vehicles, 3,000 people in the simulated population, and 400 population subgroups. The major finding is that 105 Message Calculate Audience IRPUt Message-Audience for Themes Distribution Calculate Exposure Probability Message Inputs I Calculate by Theme and Ex Time J k ' p0 Pbpulation from Stage One Ouput EXposure Statistics Go to Next Time Period Fig. 2. Kramer Mass Media Simulation Flow Diagram Stage Two: Processing Messages and Reporting Exposures SOURCE: John F. Kramer, "A Computer Simulation of .Audience Exposure in a Mass Media System: The United Nations Information Campaign in Cincinnati 1947-48" (Ph.D. dissertation, M.I.T., 1969). 106 themes "which seemed most closely related to large changes in NORC panel were also those themes for which the simu- lation predicted the highest average exposure."54 Schultz and Sullivan describe three problems with the Simulmatics model, namely the construction of a rep- “‘ resentative hypothetical population, determination of media habits, and last is the determination of the r relationship between media exposure and advertising ‘ exposure.55 Gensch created the AD-ME-SIM to remedy some t of the problems with the Simulmatics Model. AD-ME-SIM is a computer simulation media allocation model that runs in two stages like Kramer's model, the first stage con- sisting of the generation of data on the population's viewing habits which is referred to as the data gener- ation stage, and the second stage consists of the identification of the target population and is referred to as the individual weighting stage. There is, of course, a final stage which uses the individual weights generated in the first two stages and is referred to as the media evaluation stage. The limitations of the Simulmatics model are pointed out when the requirements for the input data on reading and viewing patterns are examined. The data requirements for the model are: l. The demographic, reading, and viewing habits must all come from the same individual. 2. The data must come from real individuals, not hypothetical or imaginary individuals. 107 3. The sample must be large enough to be signifi- cant on a national level. 4. The cost of gathering these data must be low enough to fall within the advertising agencies' budget.S6 Data suggested as input to this model could come from sources such as W. R. Simmons Associates, who provide relatively complete data on between 16,000 and 20,000 individuals. In the final or media evaluation stage the user inputs a proposed media plan schedule, a set of weights for the effectiveness of different media, a set of weights for the effectiveness of varying size and color ad forms, a set of weights showing the value of different patterns of exposure frequency, a list of media discounts, a set of weights showing the value of an exposure to different types of persons in the target population, and data showing the reading and viewing patterns over time of a real sample of individuals. Output from the AD-ME-SIM model includes both weekly and cumulative numbers and percentages of people in the target popu- lation reached by the proposed media schedule and the cost, the frequency, the number of exposures, but adjusted according to the subjective evaluation of the media and the value of exposures to different members of the target audience.57 Another advertising media allocation computer simulation model is the MEDIAC system reported by Little and Lodish. MEDIAC consists of an on-line 108 computer system which selects and schedules media. It consists of a market response model, a heuristic search routine, and a conversational input/output program. The user supplies various media options, a budget, various objective and subjective data; and the system selects options and schedules over time to maximize market response. Although this system contains mathemati- cal optimization routines, it is still a simulation model because of the heuristic search routine and the users' ability to experiment with various media options. Hartman and Walsh have reported a computer simu- lation of newspaper readership. The input for this model is socio-demographic characteristics from a small Iowa community, messages about National Defense, and newspaper reading habits. The simulation consists of a triple filter system. The first filter is for individuals who probably received a newspaper on the day being simulated, the second filter screens out all who had not seen the article, and the third filter screens out those indi- 'viduals who have not read the studied article. The Inajor finding of the simulation was that social status ‘variables improved prediction of readership, as well as becoming more "deterministic," i.e. excluding randomness. 'This latter finding is no doubt attributable to the fact that the study consisted of only 163 individuals.59 109 SINDI l is a diffusion model develOped by .nneman and Carroll. SINDI 1 (Simulation of Innovation ,ffusion) is a stochastic simulation model of innovation . cliques in any small, relatively closed system. The ’stem requires as input, the number of cliques, the .mber of members in each clique, and the number of 'tential tellers in each clique, the number of contacts lowed to external sources, the number of contacts Klowed to a teller once he becomes a knower, the proba- lity of a nonknower becoming a knower from an external ~urce, and the probability of becoming a knower from ~urces outside the clique. This model was able to mulate the diffusion of information about 2,4D weed ray in a Colombian peasant village.60 In addition to these mass media related simu- tions there are a number of other computer simulations at also deserve some brief mention. There are numerous mputer simulation marketing games for use as teaching vices. Among the more well-known marketing games e the M.I.T. Marketing Game which is based around e market for electric floor polishers for household e, and the Carnegie Tech Management Game special rsion called MATE (Marketing Analysis Training 61 ercise) based on a package goods industry. The NMAR media buying game developed by Schultz, Block, 110 and Jacobowitz provides a realistic media buying situation for an airline flying in the Texas triangle.62 A good example of the application of computer simulation experimentation in physical distribution research is found in the companion work of Speh and Wagenhiem. Using the LREPS simulation model, which is a model of a physical distribution system using dynamic simulation to evaluate the cost and service of alterna- tive physical distribution system designs, Speh experi- mented with demand uncertainty63 and Wagenhiem experimented with lead time uncertainty.64 More examples of computer simulation and related techniques will be discussed in Chapter IV. CHAPTER I I I--NOTES 1James S. Bright and Milton E. F. Schoeman, eds., A Guide to Practical Technological Forecasting (Engle- wood Cliffs, N.J.: Prentice-Hall, 1973), p. x. 2Robert U. Ayres, Technological Forecasting and LongrRange Planning (New York: McGraw-Hill, 1969), p. 3. 3Joseph P. Martino, Technological Forecasting for Decision-making (New York: American Elsevier, 1972), p. 2. 4Ayres, Technological Forecasting, p. x. SIbid., p. 4. 6 Klaus P. Heiss, Klaus Knorr, and Oskar Morgen- stern, Logg Term Projections of Power: Political, Economic, and Military Forecasting (Cambridge: Ballinger, T973)! PP. 3‘4. 71bid.. pp. 54-55. 8Marvin A. Jolson and Gerald L. Rossow, "The Delphi Process in Marketing Decision Making," Journal of Marketing Research 8 (November 1971); 447. 9Martino, Technological Forecasting, pp. 18-20. 10Ayres, Technological Forecasting, p. 160. 11Heiss, Knorr, and Morgenstern, Long Term Pro- jection, p. 14. 12 Ayres, Technological Forecasting, p. 177. 111 112 13Martino, Technological Forecasting, p. 130. l4Heiss, Knorr, and Morgenstern, Long Term Pro- jections, p. 65. 15Ayres, Technological Forecasting, pp. 94-95. 16Heiss, Knorr, and Morgenstern, Long Term Pro- jections, p. 67. 17 Martino, Technological Forecastigg, p. 167. laIbid., p. 205. 19Donella H. Meadows, Dennis L. Meadows, Jorgen Randers, and William W. Behrens III, The Limits to Growth (New York: Universe Books, 1972). 20Martino, Technological Forecasting, p. 205. 21Mark L. Hinshaw, "Wiring Megapolis: Two Scenarios," in Communication Technology and Social Polic , eds. G.’Gerbner, L. Gross, and W. Melody (New York: John Wiley & Sons, 1973), pp. 307-16. 22Paul Baran, "Some Changes in Information Tech- nology Affecting Marketing in the Year 2000," in Changing Marketing Systems, ed. R. Moyer (Chicago: American Marketing Association, 1967), p. 82. 23E. Bryan Carne, "Telecommunication: Its Impact on Business," Harvard Business Review 50 (July-August 1972): 125. 24H. J. Barnett and E. Greenberg, "On the Eco- nomics of Wired City Television," American Economic Review 58 (June 1968): 504-05. 251bid., p. 506. 26Kan-Hua Young, "The Demand and Supply of Tele- communication Services in the United States: Empirical Results," in gong Term Projections of Power, eds. K. Heiss, K. Knorr, and O. Morgenstern (Cambridge: Ballinger, 1973), p. 203. 113 27James C. Ohls, "Marginal Cost Pricing, Invest- ment Theory and CATV," The Journal of Law and Economics 13 (1970): 459-60. 28Rolla Edward Park, Potential Impact of Cable Growth on Television BroadcastingiTSanta Monica: Rand Corporation, 19707) pp. 8-11. 29R011a Edward Park, "The Growth of Cable TV and Its Probable Impact on Over-The-Air Broadcasting," American Economic Review 61 (May 1971): 73. 30Robert P. Abelson, "Simulation of Social Behavior,” in The Handbook of Social Psychology, Vol. II, eds. G. Lindzey and E. Aronson (Reading: Addison-Wesley, 1968), p. 274. 31Ithiel de Sola Pool, "Computer Simulation of Total Societies," in The_§tudy of Total Societies, ed. S. Klausner (New York: Frederick A. Praeger, 1967), p. 52. 32F. Gerald Kline, "Theory in Mass Communication Research," in Current Perspectives in Mass Communi- cation Research, Vol. I, eds. F. Kline and P. Tidhenor (Beverly Hills: Sage, 1972): p. 26. 33Randall L. Schultz and Edward M. Sullivan, "Developments in Simulation in Social and Administrative Science," in Simulation in Sogial and Administrative Science, eds. H. Guetzkow, P. Kotler, and R. Schultz (EngIewood Cliffs, N.J.: Prentice-Hall, 1972), p. 3. 34Ibid., p. 4. 35Abelson, "Simulation of Social Behavior," p. 275. 36 p. 57. Pool, "Computer Simulation of Total Societies," 37John M. Dutton and Warren G. Briggs, "Simu- lation Model Construction," in Computer Simulation of Human Behavior, eds. J. Dutton and W. Starbuck (New York: John Wiley & Sons, 1971), p. 104. ——;___~ud 114 38Abelson, "Simulation of Social Behavior," p. 284. 391bid., pp. 284-85. 40William T. Newell and Robert C. Meier, "Business Systems Simulation," in Simulation in Social and Admin- istrative Science, eds. H. Guetzkow, P. Kotler, and R. Schultz (Englewood Cliffs, N.J.: Prentice-Hall, 1972), p. 447. 41Ithiel de Sola Pool, Robert P. Abelson, and Samuel L. Popkin, Candidates, Issues and Strategies: A Computer Simulation of the 1260 and 1964 Presidential Elections (Cambridge: M.I.T. Press, 1965), pp. 102-03. 42Schultz and Sullivan, "Developments in Simu- lation," pp. 15-16. 43Ibido I pp. 28-30. 44Newell and Meier, "Business Systems Simu- lation," pp. 416-32. 45Philip Kotler and Randall L. Schultz, "Market- ing Systems Simulation," in Simulation in Sogial and Administrative Sgience, eds. H. Guetzkow, P. Kotler, andiR. Schultz (Englewood Cliffs, N.J.: Prentice-Hall, 1972). pp. 482-83. 46Ibid., p. 544. 47Abelson, "Simulation of Social Behavior," p. 281. 48Thomas H. Naylor, Joseph L. Balintfy, Donald S. Burdick, and Kong Chu, Computer Simulation Techniques (New York: John Wiley & Sons, 1966), pp. 36-37. 491bid.. pp. 26-41. 50Thomas H. Naylor, "Methodological Consider- ations in Simulating Social and Administrative Systems," in Simulation in Social and Administrative Science, eds. H. Guetzkow, P. Kotler, and R. Schultz (Englewood Cliffs, N.J.: Prentice-Hall, 1972), pp. 662-65. 115 511bid.. pp. 654-56. 52Schultz and Sullivan, "Developments in Simu- lation, I! ppe 517-180 ' 53Samuel L. Popkin, "A Model of a Communication System," American Behavioral Scientist 8 (1965): 8-11. 54John F. Kramer, "A Computer Simulation of Audience Exposure in a Mass Media System: The United Nations Information Campaign in Cincinnati 1947-48" (Ph.D. dissertation, M.I.T., 1969). 55Schultz and Sullivan, "Developments in Simu- lation," pp. 519-20. 56Dennis H. Gensch, "A Computer Simulation Model for Selecting Advertising Schedules," Journal of Market- ing Research 6 (May 1969): 204. 57Ibid.. pp. 209-10. 58John D. C. Little and Leonard M. Lodish, "A Media Planning Calculus," Operations Research 17 (January-February 1969): l. 59John J. Hartman and James A. Walsh, "Simulation of Newspaper Readership: An Exploratory Analysis of Social Data," Social Science ggarterly 49 (1969): 840-52. 6oGerhard J. Hanneman, "A Computer Simulation of Information Diffusion in a Peasant Community" (Master's thesis, Michigan State University, 1969). 61Kotler and Schultz, "Marketing Systems Simu- lation," pp. 537-38. 62Don E. Schultz, Martin P. Block, and Bret Jacobo- witz, DONMAR I: A Computer_Simu1ated Media Buying Game, Michigan State University, East Lansing, Michigan, 1975. 63Thomas W. Speh, "The Performance of a Physical Distribution Channel System Under Various Conditions of Demand Uncertainty: A Simulation Experiment" (Ph.D. dissertation, Michigan State University, 1974). 116 64George D. Wagenheim, "The Performance of a Physical Distribution Channel System Under Various Con- ditions of Lead Time Uncertainty: A Simulation Experi- ment" (Ph.D. dissertation, Michigan State University, 1974). O; at C3 0 n f0 a1. ”he CHAPTER IV TIME ALLOCATION STUDIES Time as a Social Variable Moore describes time and space as a way of locating human behavior, or "a mode of fixing the action that is peculiarly appropriate to circumstances."1 All of human behavior can be fixed in a point in both space and time, including communication behavior. Communi- cation involves sending messages through space or dis- tance, as well as devoting time to the process. Communi- cation can be measured in terms of the manipulation of both space or distance, as well as time. While Space and time are interdependent in fixing human behavior, it is possible to consider them separately for theoretical purposes. Models considering only the allocation or manipulation of space are considered static, whereas those models considering the allocation or manipu- lation of time are considered dynamic. As Moore points out: Short of major and sudden geophysical events, space principally acquires any dynamic qualities it may have by virtue of changes in social values, interests and techniques. In other words, space is generally 117 118 a passive condition of behavior, variable only as human behavior makes it so. Time, on the other hand, is intrinsically dynamic, and indeed the idea of dynamic (or static) is impossible without reference to conceptions of time. Human communication activity can be described in terms of two finite resources, space and time. Spatial models consider the arrangement of the communication activity in space and provide a static picture of the process in terms of distance and potential of communication flow. Temporal models consider the allocation of available time to communication activity and provide a dynamic analysis of the communication process. As Moore describes, "the value of time tends to be judged in terms of its use. . . . "3 Communication need, as defined by Reid, can be viewed as a trade-off between the allocation of space and the allocation of time. Models of the allocation of space and distance in social settings generally take the form of a network flow model. Meier proposes a communication theory of urban growth by pointing out that intensification of communications, knowledge, and controls, are highly correlated with the growth of urban centers. Meier explains that the theory begins considering the "utility of agglomeration" for human beings. Human beings require some form of social cohesion for the purpose of genetics, establishing primary relationships necessary for nuture, formation of coalitions for 119 purposes of defense, and formation of coalitions for the exchange of goods. This leads to the accumulation of knowledge and to the selection and recombination of the most useful elements of culture. Technological innovation, such as writing, the invention of the print- ing press, and the electronic media have caused spurts in the accumulation of knowledge.4 BCN technology might be expected to contribute to an additional spurt. While space and distance are important determi- nants of human activity, sufficient empirical data do not exist to allow physical distance to become part of a human activity model. An exception is the inclusion of mean locus or the distance for out-of—home activities reported by Chapin.5 The mean locus is as the "crow flies" distance between points such as home and work. Typically human activity studies, or time allocation studies, do not include any measurement of physical distance, but rather consider the problem of movement through physical space as a problem of travel time. While recognizing the importance of spatial measures, it is necessary to concentrate on time. Heirich describes four distinct ways time can be used in the study of social change. Time can be used as a social factor in the explanation of changes, as a causal factor between other elements, as a quanti- tative measure of them, or as a qualitative measure of 120 their interplay.6 Time as a social factor influences social interaction as a resource and as social meaning. Time as a resource refers to the time alloted to various activities and is the View of time held in most time- budget research. Time as social meaning refers to spe- cific moments in time such as Christmas or time sequence which may indicate priority, such as "work before play." Time as a causal link can be viewed in terms of setting, or space-time relationships, and sequence, or time-time relationships. While time itself is not sufficient to establish causal relationships, it provides the backdrop. Time can provide a measure of the quantitative changes in a relationship such as rate or speed, and duration. Time can also provide a qualitative measure of social change such as a qualitatively different social structure. Time is important to allow the interplay of various social forces which ultimately contribute to a social change.7 Economists consider time as a scarce resource. Linder summarizes the position as follows: The analysis of the distribution of time, of changes in this distribution arising from economic growth, and of the implications of economic development under an increasing scarcity of time is not something of purely economic interest. It is rather a problem of more general interest, a joint problem for all the social sciences. The distribution of time and changes in this distribution are bound to affect our entire attitude to social problems, our entire philosophical outlook.8 ' 121 Linder argues that consumption should be viewed not only as a problem involving physical goods, but also as a time problem. Consumption involves both the time necessary to consume the goods as well as the goods themselves. Consumption should not be viewed as an instantaneous process, the time at a consumer's disposal must also be considered. In examining economic growth, Linder categorizes time into five separate groups which he believes to be "unequal from a philOSOphical point of view." The first category is working time, or the time spent working in specialized production. WOrking time has impact on both the supply and demand for time to allocate to other activities. The second category is personal work, or services, which can be divided into the maintenance of goods and of one's body. A third category is con- sumption time, or the time allocated to the consumption of goods. As productivity increases, so does the demand for consumption time. A fourth category is time devoted to the cultivation of mind and spirit. The primary difference between this category of time and consumption time is the role that goods play. Obviously the cultivation of mind and spirit is less affected by consumer goods, that is consumption time. The fifth and final category of time is free time or idle time.9 122 It is the last category of time that has received considerable attention under the rubric of leisure time. An example is Foote's methodological essay on the application of time-budget studies to problems of the aging.10 Of more interest here is the use of time as a variable in marketing. Time as a Marketing Variable According to Schary, far more is known about how consumers spend their money than how they spend their time.11 Time has been largely ignored in marketing because it has not been part of the accepted framework of consumer behavior. While it has not been the center of attention, models of time allocation have evolved in the traditional economic and consumer behavior literature. Marketing related time allocation models can be divided roughly into two groups: those that assume man to be rational and those that do not. Because of the origin of most of the models in economics, the majority assume man to be rational. Although the difference may be moot, as Bucklin concluded, that for the most part, the economic theory of consumer behavior is supported by the findings of his research as discussed in Chapter II. In particular, Bucklin adds, the generalization that consumers would respond to lower shopping costs by making in-store comparisons is especially strong.12 ’123 Several rational man time allocation models using neoclassical economic consumer theory have been developed. Becker Opened the field with his model evaluating the effect of foregone earnings on leisure time. Addressing himself to the reduction in average working time which occurs along with economic develop- ment, Becker makes the point that allocation and effi- ciency of nonworking time is becoming more important. Becker summarizes the model as follows: At the heart of the theory is the assumption that households are producers as well as consumers; they produce commodities by combining inputs of goods and time according to the cost-minimization rules of the traditional theory of the firm. Commodities are produced in quantities determined by maximizing a utility function of the commodity set subject to prices and a constraint on resources. Resources are measured by what is called full income, which is the sum of money income and that foregone or "lost" by the use of time and goods to obtain utility, while com- modity prices are measured by the sum of the costs of their goods and time inputs.13 Johnson criticizes time allocation models, such as Becker's, that assume the opportunity cost of an hour of leisure is equal to the money wage rate earned by the individual. Johnson proposes a model concentrating on the problem of travel time that assumes that indi- vidual behavior is subject to a time budget constraint as well as a money budget constraint, and that work and leisure are distinct decision variables in the utility function.14 Models that assume the money wage rate tend to overstate the value of leisure and travel time. 124 Mabry suggests an even more complex set of multiple constraints on the choice of nonwork activity, including time, stamina, and money.15 DeSerpa has developed a general theory of the economics of time without any focus on specific problems such as the effect of foregone earnings upon consumer choice, and the valuation of travel time. The essential features of the model include utility represented as a function of both commodities and the time allocated to them, individual decision-making is subject to both a money constraint and a time constraint, and the decision to consume a commodity requires that some minimum amount 16 The economic models of of time be allocated to it. time allocation provide a basis for evaluating the importance of the fundamental trade-off between com- munication and transportation, and between alternate forms of communication. The difficulty in applying the economic models to the BCN technology problem is the lack of empirical specification, that is the models are neoclassical mathematical optimization models. Other approaches to consumer behavior are char- acterized by studies of the convenience-oriented con- sumer. Mauser predicts, for example, that the affluent citizen of the next century will be oriented "to buying 17 time rather than product." Rathmell describes the newly acquired American freedoms of discretionary time 125 and discretionary mobility.18 Rathmell believes that discretionary mobility along with the necessary free time releases the consumer from habitual shopping patterns. Consumers will begin to classify shopping as an enjoyable activity, rather than as a routine household chore.19 Anderson's study, discussed in Chapter II, describes distinct consumer typologies of convenience good orientations. Determinants of con- venience orientation include socio-economic status and stage in the family life cycle.20 Considering the use or allocation of time as an indication of life-style is suggested by Andreasen. Life-style, according to Andreasen, is a social science concept connoting the totality of behaviors which com- prise the characteristic approach to life of a particular individual or group.21 Andreasen suggests the use of time-budgets, or simply the amount of time an individual allocates to various activities through the day, as a ‘way'of measuring life-style. Comparing changes in time- budgets, life-style comparisons across different occu- jpational groups, as well as predictions about future time ‘use can be made. Interestingly, increases in leisure activity, especially away from home, are predicted by Auuireasen along with a decline in in-home activities such.ae television viewing.22 Andreasen bases prediction on an analysis of the United States Time Use Survey 126 which is the basis for the computer simulation model which is described in Chapter V. Time-budget analysis provides an insight into the future as well as a measure of life-style. Schary provides an outline of a framework for the consideration of the ways consumer time choice influences marketing. The framework recognizes the different roles that time plays in the selection of work and leisure activities. The nature of the activity chosen will also influence the role of time. For example, if the activity can overlap with other activity, or whether or not the activity is a solo activity or is done in a social setting, as well as the relationship of the activity to other activity. It is common to engage in another activity while listening to the radio; and making a shopping trip often requires travel. Time values also follow a cyclic pattern, and violation of the customary pattern would be expected to create feelings of time urgency beyond those in the normal situation. ‘Underlying all of the classifications of time allocation are:the subjective attitudes held by the participating individuals or groups in a particular situation or actiyity.23 Schary concludes with the following point: Time is an implicit part of every market offering. Products are chosen not only because of price, quality, features, or even the images created through promotional activity, but also because of the potential time expenditures that they entail. Therefore, time should become an explicit element in marketing strategy.24 127 Time-Budget Studies According to Szalai, it is not known when, where, or by whom the internationally adopted term "Zeitbudget" or time-budget was first used in the technical literature.25 A time-budget study is for the most part a methodology employed by sociologists and generally involve large-scale survey research. A definition of time-budget research is provided by Szalai: Many interesting patterns of social life are associated with the temporal distribution of human activities, with regularities in their timing, duration, frequency and sequential order. Certain techniques of data collection based on direct observation, interviewing and the exami- nation of records permit the establishment of fairly adequate itemized and measured accounts of how peOple spend their time within the bounds of a working day, a week-end, a seven-day week, or any relevant time period. Investigations of this particular aspect of social life based on the quantitative analysis of such accounts are commonly referred to as "time-budget studies."26 Time-budget studies generally go farther than the description of the timing, duration, frequency, and sequential order of various human activities throughout the day; and almost always involve comparison of various groups of individuals in order to make inferences about the groups. The application of the large-scale time-budget studies, however, have been somewhat limited in any ability to explain time allocation. Taking a critical View of time-budget studies, Linder writes as follows: 128 The sociologists, for their part, have made great efforts to perform large-scale time budget studies. They have tried to plot how different individuals or groups divide up their time between various activities. Particularly detailed studies have been made of the use of time spent outside the place of work, time which is devoted to a variety of different activities. However, the theories formed parallel to these studies have been of an ad hoc nature. Attempts at any systematic expla- nation of time allocation and changes in it are lacking.27 Linder, an economist, believes that the results of these time-budget surveys have never really been used, because the importance of time scarcity in the economic sense has been ignored. Linder concedes that if an analysis could yield a dynamic theory capable of making socio- logical predictions, it could become a useful tool in the study of the future.28 The computer simulation model described in Chapter V is an attempt, albeit some- what crude, to develop such a tool for the study of the future. Historically time-budget studies grew out of family expenditure studies in England and France in the late eighteenth century. Chapin cites Sir Frederick Morton Eden's State of the Poor published in 1797 as one of the first known works to make use of family 29 Szalai describes Frederick budget information. Leplay's use of family budget information in his study of the living conditions of the working class in various European countries in the mid-nineteenth century. Szalai also describes Eduard Engel's Das Rechnungsbuch 129 der Hausfrau und seine Bedeutung im Wirtshaftsleban der Nation published in 1882. Engel drew attention to various regularities in family budgets and formulated his laws. One of the most well-known of Engel's Laws is that as income increases in the lowest categories of family income the greater part is spent on food. Upon further rise in the standard of living, increases in income are spent proportionately less on food, and more on other things.30 Time-budget studies continued in the Soviet Union since the very beginning of the communist government and were viewed as important tools in economic planning.31 In the United States some of the earliest experience in using time allocation came from the "time and motion" studies of factory management by Frederick W. Taylor done in the very early part of the twentieth century.32 A number of studies with a sociological orientation followed, including studies of farm house- wife's use of time and the use of leisure in suburban areas. The most well-known of the early studies is the Sorokin and Berger study of the Boston area published in 1939.33 The Sorokin and Berger study is the source of the 1935 time-budget data cited in Chapter I. The Sorokin and Berger study is considered the forerunner of the contemporary time-budget studies. The problems addressed by Sorokin and Berger in 1939 130 are quite similar to those of current day time-budget studies and were expressed as follows: 1. What kind of activities occupy the individual's twenty-four hours? How often is each activity repeated during the twenty-four hours? How much time does each activity take? What are the individual motives for each activity? Does the activity have one or several motives? Does the activity have the same or different motives with different individuals? And, vice versa, does the motive manifest itself in identi- cal or different activities with different indi- viduals, or with the same individual at different hours of the day? What part of the twenty-four hours does an indi- vidual spend alone and what part with another individual or with groups? Who are these persons and groups? What activities cause the individual to spend his time with each of these? What is the length of time he lives and acts in associ- ation with each of these? How accurately can an individual predict his behavior twenty-four hours, forty-eight hours, one week, and one month in advance? Finally, what are the differences, if any, in all of the above respects between individuals of different sexes and different ages? Are there any tangible and uniform variations in the above aspects of human behavior on different days of the week?34 Szalai summarizes the topics covered in the early time-budget studies as follows: 1. 2. 3. the share that such broad categories of activity like paid work, housework, personal care, family tasks, sleep and recreation have in the daily, weekly or yearly time-budget of the population; characteristic time expenditures of certain social groups or strata on more or less specified types of everyday activities: the use made of "free time," leisure. especially: Contemporary use of sociological time-budget studies include the analysis of activity patterns as demonstrated 131 by Brail and Chapin,36 analysis of social patterns as demonstrated by Schneider,37 analysis of mass media con- sumption as demonstrated by Robinson and Converse,38 and analysis of technological progress as demonstrated by Staikov.39 The research strategy employed in sociological time-budget studies defines the activity patterns as dependent variables and certain preconditioning and predisposing factors as independent variables. An activity pattern generally refers to a tendency for people in a given population to behave in a certain way. It has the properties of duration, position in time usually designated by start time, a place in a sequence of events, and a fixed location in space. Activity patterns are expressed in terms of an activity classifi- cation scheme.40 Chapin provides an example of an activity class with "shopping." A problem with time- budget studies is directly related to activity classifi- cation schemes. For example, if the concern of the study is with shopping as a phenomenon of the culture, then the shopping classification is probably adequate. If the concern is with something like public transpor- tation planning, then shOpping would have to be redefined to include such activities as driving from the home to the shopping center, buying groceries, then driving home again.41 The development of an appropriate 132 activity classification system is one of the major methodological issues in time-budget research. The independent variables, or preconditioning and predisposing factors, are divided into two groups by Chapin. The first group consists of subsistence needs and the second group consists of culturally, socially, and individually defined needs. Chapin out- lines these factors as follows: 1. Subsistence needs a. Basis of motivation: need for sleep, food, shelter, clothing and health care. b. Requisite means of satisfying needs: for example, income-earning Opportunities from vocational training, education, medical care, social service, etc. 2. Culturally, socially, and individually defined needs a. Basis of motivation: felt needs for security, status, achievement, affection, and social contact; outlets for exercise of personal talents, ingenuity, prowess, and skill; need for mental release, for example, the release of feelings of joy, fear, frustration, or alienation; and need for physical release, for example, physical exercise as well as rest and relaxation. b. Requisite means of satisfying needs: oppor- tunities for seeing kinsmen, friends, neighbors, and others; opportunities for participation in church, voluntary organi- zations, and civic activities; opportunitieS' for creative activity, for engaging in recre- ation and other diversions, and for rest and relaxation.42 The resultant research strategy, then, is to postulate varying combinations of these motivations in explaining activity choice. Time-budget studies have been used to predict the future. Holman predicts a tripling in the amount 133 of discretionary time available to Americans from that available in 1950 by the year 2000. The prediction is based upon a secondary analysis of a number of published time-budget studies including Sorokin and Berger.43 Although, it should be pointed out, the Holman analysis does not include the development of any mathematical model. Staikov suggests that time-budget studies can provide a way of measuring technological progress. The social value of technological progress, according to Staikov, depends on the total amount of time which can be saved by use of the technology. Staikov also makes the point that technological progress can only be under- stood by examining all of man's activities. By intro- ducing a new technology, time spent in one activity category may be decreased, but then there must be an increase in other types of activities.44 The use of time-budget information to evaluate technological pro- gress, then, must be comprehensive with respect to all categories of activity. There are a number of contemporary large-scale time-budget studies which could become the empirical base for the development of a computer simulation model to evaluate the impact of BCN technology. The Multi- national Comparative Time-Budget Research Project was organized in 1963 and includes social scientists and 134 time-budget projects from Belgium, Bulgaria, Czechoslo- vakia, France, East and West Germany, Hungary, Peru, Poland, U.S.A., U.S.S.R., and Yugoslavia. The project was organized around the following aims: 1. To study and to compare in different societies variations in the nature and temporal distribution of the daily activities of urban suburban popu- lations subjected in varying degrees to the influences and consequences of urbanization and industrialization; 2. To develop methods and standards for the collection and evaluation of data pertinent to temporal and other dimensions of everyday activity which, apart from their interest to social theory, are also of considerable importance for the organi- zation of working life and for the creation of satisfactory conditions for the enjoyment of leisure; ‘ 3. To establish a body of multinational survey data on characteristics of everyday life in urban surroundings under different socio-economic and cultural conditions which could serve as the basis for testing various methods and hypotheses of cross-national comparative social research. 4. To promote, in general, cooperation, standardi- zation of research techniques and the exchange of quantitative data at an international level among social scientists involved in survey research who are endeavoring to achieve com- parable results with a view to their common evaluation. ' Included in the Multinational Project are two studies from the United States done by Converse and Robinson at the Survey Research Center of the University of Michigan. One study was conducted in Jackson, Michigan consisting (of 778 respondents, and one study was using a national jprobability cross-section drawn from forty-four metro- ;politan areas. Both studies were conducted between iNoyember 1965 and April 1966.46 The later study, 135 referred to as the Forty-four Cities, U.S.A., was selected as the empirical base for the development of the simulation model and will be discussed in more detail. Other time-budget studies include the family of activity studies undertaken primarily by the Center for Urban and Regional Studies at the University of North Carolina. Brail and Chapin report two national surveys from samples drawn from forty-three Standard Metropolitan Statistical Areas. The first, done in 1966, consisted of 1,467 households. The second, done in 1969, was a follow-up of 1,199 of the 1966 households. Since both studies were not centrally concerned with daily activity, but rather residential mobility, questions relating to activity had to be kept to a minimum. In both studies respondents were asked to reconstruct yesterday's nonpersonal activities from the time they got up until the time they retired.47 Chapin reports a series of studies in Washington, D.C. beginning in 1968. The basic study had a primary focus on household use of medical care services and facilities, but obtained substantial activity information from 1,756 respondents. Two additional studies were done in 1969 among the inner-city black community and in 1971 among the tran- sitional white community.48 The Forty-Four City study was selected as the Ibasis for this project because of the generalizability 136 of the national sample, and because of the richness of the data reported in the literature. The reports of the Brail and Chapin, and Chapin studies do not include the volume of data reported in the Forty-Four City study, or as titled by the Survey Research Center, the United States Time Use Survey. The Forty-Four City study is the only available time-budget data source that provides in any form the data necessary to estimate the proba- bility distributions necessary for the development of the computer simulation model described in Chapter V. United States Time Use Survey The United States Time Use Survey conducted by the Survey Research Center of the University of Michigan had four major goals. The first was the collection and comparison of basic behavioral data across eleven nations as part of the Multinational Comparative Time-Budget Project. Second was the tabulation of general descrip- tive information about life in the United States. Third was the use of these data as bench-marks in the measurement and assessment of social change. Fourth and last was the investigation into the major activities and objects which bring gratification and satisfaction to individuals in different parts of society.49 The survey was conducted using the regular con- tinuing national sample of the Survey Research Center chosen to represent the United States. A total of 137 forty-four metropolitan areas are included in the sample designed to secure 1,500 respondents. Characteristics of the sample are shown in Table 2. Due to an unexpectedly high proportion of households without an eligible respondent, that is households where all respondents were over age 65, the total national sample came to 1,244 respondents.50. For a more detailed description of the sample and the metropolitan areas refer to the Survey Research Center report.51 The interviewing procedure consisted of a recruitment personal interview leaving a diary for the respondent to record a single day's activities. A follow-up personal interview was then conducted at least one day later to help complete the diaries and to collect them. The first personal interview was short, about twenty to twenty-five minutes, and the second interview was somewhat longer lasting forty to fifty minutes. Diaries were distributed so that all days of the week would be represented. The field work was conducted in two waves. The first wave was con- ducted between November 15 and December 15, 1965, and yielded 936 usable nonfarm employed respondents. A second "clean-up" wave added 308 additional respondents and was conducted between March 1 and April 25, 1966. An overall completion rate for both waves of 81.5 percent was obtained.52 138 TABLE 2 SUMMARY CHARACTERISTICS OF THE NATIONAL TIME USE SURVEY SAMPLE The (a) (b) (c) (d) (e) (f) (g) (h) (i) (j) (k) (1) final national urban sample contained: 543 men, 701 women 345 people under 30, 283 aged 30-39, 305 aged 40-49 and 304 aged 50 or over 357 housewives, 342 employed women, 519 employed men, 13 students and 13 nonworkers 40 unskilled, 365 (semi) skilled, l8 technical, 189 lower white-collar, 94 upper white-collar and 172 professional workers 203 workers were employed in manufacturing industries, 240 in trade and retail, 107 in technical public ser- vice, 81 in education, 52 in construction, 61 in administrative public service, and 136 in the remain- ing other branches of the economy 191 respondents had graduated from college, 647 from high school, 321 from primary school and 72 had not graduated from primary school 995 respondents were married, 127 single, 57 widowed and 62 divorced or separated 470 had no children in the household, 300 had children all over 4 years old, 460 had no child under 4 years old 922 lives in residential sections of cities, 55 in the industrial and business districts, 71 in small towns outside larger cities and 129 in isolated houses out- side of Jackson city 1,101 respondents had an automobile belonging to the household 149 respondents earned over $10,000/year, 262 between $7,500 and 9,999, 494 between $5,000 and 7,499, 205 between $3,000 and 4,999 and 100 under $3,000 (note income is for the year 1965) 745 respondents considered themselves Protestant, 383 Catholic, 62 Jewish and 39 expressed no religious preference SOURCE: Summary of United States Time Use Survey. Survey Research Center, Institute of Social Research, University of Michigan, Ann Arbor, Michigan, 1966. 139 Approximate calculation of F—ratios are shown between all the sociological groupings in the tables presented in the Survey Research Center report. Only overall "omnibus F-tests" are reported, and for more detailed comparisons the special range tests, either Tukey or Scheffé, are recommended as apprOpriate sta- tistical procedures.53 Analytical Models of Time- Budget Data Time-budget studies consist of large-scale sociological studies using survey research methodology to obtain estimates of the participation rate and the mean duration of daily activities for some population. Of critical importance to the analysis of time-budget data is the classification scheme applied to the various daily activities. The Multinational Comparative Time- Budget Project employed a two-digit 96 activity category scheme which was also used by the United States Time Use Survey. The 96 activity category scheme while accommo- dating most activity is somewhat difficult to analyze. A reduced categorization scheme of 37 activities par- tially alleviates the problem and is the standard activity categorization scheme employed in the Multi- national Project.54 Other activity classification schemes vary somewhat from the Multinational Project, Chapin, for example, reports a 225 activity classification 140 scheme which is reduced to a 40 activity scheme and a 55 The 12 activity scheme, depending upon the situation. computer simulation model developed here relies on a further reduction of the 37 activity categorization scheme because the United States Time Use Survey is the source of the data. This is discussed in more detail in Chapter V. Stone describes five types of analytical models appropriate to the analysis of time—budget data. The first of these is accounting for variance. This is the most commonly applied analytical model and conforms to the research strategy for time-budget studies described earlier. The goal is to describe either groups or motivations that produce certain patterns of activities, that is what factors account for more variance in the time-budget. The second analytical model is the Markov model. Markov chains are appropriate whenever a time sequence is involved and when one event is predicted by a preceding event. Although, as Stone points out, brief consideration of how time is allocated leads to the dis- counting of such a simplistic assumption as one event being determined by a preceding one. Stone concludes that Markov models cannot address the complexities of time-budget data. The third analytical model is eco- nomic elasticity. Elasticity models, such as those described here earlier, predict overall time allocations 141 on the basis of tradeoffs, that is more time spent on one kind of activity implies less time for another. The fourth type of analytical model discussed by Stone is path analysis of causality. Path analysis involves the developing of coefficients for estimating the depen- dency of one variable on another, within a causal analysis framework. Such techniques are well suited to time-budget problems because they allow for a variety of causal dependencies other than precedence, including elasticity.56 The most sophisticated analytical model appropriate to time-budget data discussed by Stone is the process model. A process model attempts to represent the decision making that is involved in making time allo- cations. The sequence of a process model tends to have the natural representation in the steps of a computer program, especially by the "if" statement. Stone goes on to discuss the major limitation of process models, which is the large number of variables and relationships which must be Specified to describe even a simple activity such as reading. Probably the only justifi- cation for developing models is in answering important social questions, rather than in merely being able to describe a process. An example might be in traffic and pollution. New plans could be designed as inducements to change travel habits to reduce congestion and 142 pollution. The effects of such plans could then be anticipated by running them against the computer model.57 Stone's application of the process model is the application that makes the technique attractive to the urban planner. It is this attraction that has led to the develOpment of any time-budget process or computer simulation literature that exists in the field of urban planning. Urban Planning Time-Budget Computer Simulation Models While a working time-budget simulation model has yet to be published, almost all of the literature sug- gesting the development of such a model is in urban planning. Chapin suggests a four-phased research and development effort focused on human activity systems in the metropolitan community as follows: 1. Description, that is, a study of patterned ways different sub-societal segments of the metro- politan community use the city, its facilities, 'and its services; 2. Explanation, that is, a study of the factors that appear to regulate patterns thus described; 3. Simulation, that is, the development of a model capable of reproducing activity patterns, incor- porating these explanatory factors; 4. Evaluation, in which the simulation model is used to investigate the likely impacts on human activity of the implementation of various alternative plans and policies. It is this approach that makes the computer simulation of time-budget data an attractive tool in technological forecasting. The ability to evaluate the impact of a 143 new technology using a simulation model can provide valuable information to both public and private sector planners. Although, as Hemmens points out, the develop- ment of such a model is "clearly an enormous task and, possibly, even a misguided one."59 Chapin cites only two simulation models extant in the urban planning literature, the Brail model in 1969 and the Tomlinson, Bullock, Dickens, Steadman and Taylor model in 1973.60 Tomlinson et a1. developed a quantitative model to predict the distribution of stu- dents in different activities and locations during a typical day, depending on the restrictions imposed by the spatial distribution of buildings and sites, and by administrative and social constraints on the timing of activities. They describe the model as an "entropy- maximizing type," and derived the mathematical relation- ships from two student time-budget studies. The model developed in England has two purposes, first as a research tool for examining particular types of uni- versity organization, and second to develop general methods of studying and modeling human spatial behavior.61 The model is less than perfect, however, as the authors conclude "the model requires further development and refinement before it can be applied with confidence to the solution of actual planning problems."62 Tomlinson et al., however, point out a serious problem in the 144 development and utilization of such models, and that is the stability of time allocations under different physical conditions that may be imposed through the use of the model.63 The Brail microanalytical model exists only on block diagram form and has not been published in any further stage of development. The Brail model is the basis for the computer simulation model developed here and is discussed extensively in Chapter V. CHAPTER IV--NOTES lWilbert E. Moore, Man, Time, and Society (New York: John Wiley & Sons, 1963), p. 7. 2Ibid., p. 8. 31bid., p. 19. 4Richard L. Meier, A Communication Theory of Urban Growth (Cambridge: The M.I.T. Press, 1962). 5F. Stuart Chapin, Jr., Human Activity Patterns in the City (New York: John Wiley & Sons, 1974), p. 71. 6Max Heirich, "The Use of Time in the Study of Social Change," American Sociological Review 29 (June 1964): 386. 7Ibid.: pp. 387-89. 8Staffan Burenstam Linder, The Harried Leisure Class (New York: Columbia University Press, 1970), pp. 3-60 9 Ibide , pp. 13-150 10Nelson N. Foote, "Methods for Study of Meaning in Use of Time," in Aging and Leisure: A Research Prospective into the Meaningful Use of Time, ed. R. Kleemeier (New York: Oxford University Press, 1961), pp. 155-76. 11Philip B. Schary, "Consumption and the Problem of Time," Journal of Marketing 35 (April 1971): 51. 145 146 12Louis P. Bucklin, "Testing Propensities to Shop," Journal of Marketing 30 (January 1966): 26. 13Gary S. Becker, "A Theory of the Allocation of Time," Economic Journal 75 (September 1965): 516. l4M. Bruce Johnson, "Travel Time and the Price of Leisure," Western Economic Journal (Spring 1966): 135. 15Bevars D. Mabry, "An Analysis of Work and Other Constraints on Choices of Activities," Western Economic Journal 8 (September 1970): 214. 16A. C. DeSerpa, "A Theory of the Economics of Time," The Economic Journal 81 (December 1971): 828. 17Ferdinand F. Mauser, "A Universe-In-Motion Approach to Marketing,” in Managerial Marketing: Per- 5 ectives and Vieypoints, eds. E. Kelley and W. Lazer Homewood, 111.: Richard D. Irwin, 1967), p. 50. 18John M. Rathmell, "Discretionary Time and Dis- cretionary Mobility," in Managerial Marketing: Per- gpectives and Viewpoints, eds. E. Kelley and W. Lazer Homewood, III.: RiChard D. Irwin, 1967). p. 147. 191bid., p. 154. 20W. Thomas Anderson, Jr., "Identifying the Convenience-Oriented Consumer," Journal of Marketigg Research 8 (May 1971): 179-83. 21Alan R. Andreasen, "Leisure, Mobility, and Life-Style Patterns," in Changing Marketing Systems, ed. R. Moyer (Chicago: American Mafketing Association, 1967), p. 55. 221bid., p. 62. 23Schary, "Consumption and the Problem of Time," pp. 54-55. 24 Ibid., p. 55. 147 25Alexander Szalai, "Trends in Comparative Time-Budget Research," The American Behavioral Scientist, December, 1966, p. 3. 26Alexander Szalai, ed., The Use of Time (The Hague: Mouton, 1972), p. 1. 27Linder, The Harried Leisure Class, p. 6. 28Ibid. 29Chapin, Human Activity Patterns, p. 3. 30Szalai, "Trends in Time-Budget Research," p. 3. 3lIbid., p. 4. 32 . . . Chapin, Human ActiVlty Patterns, p. 4. 33Ibid. 34Pitirim A. Sorokin and Clarence Q. Berger, Time-Budgets of Human Behavior (Cambridge: Harvard University Press, 193977'pp. 3-4. 35Szalai, The Use of Time, p. 6. 36Richard K. Brail and F. Stuart Chapin, Jr., "Activity Patterns of Urban Residents," Environment and Behavior 5 (June 1973): 163-90. .1 37Annerose Schneider, "Patterns of Social Inter- action," in The Use of Time, ed. A. Szalai (The Hague: Mouton, 1972), pp. 317-34. 38John P. Robinson and Philip E. Converse, "The Impact of Television on Mass Media Usages: A Cross- ‘National Comparison," in The Use of Time, ed. A. Szalai (The Hague: Mouton, 1972). PP. 197-212. 39Zahari Staikov, "Time-Budgets and Technological Progress," in The Use of Time, ed. A. Szalai (The Hague: Mouton, 1972), pp. 461-82. 148 4OChapin, Human Activity Patterns, p. 37. 41Ibid., p. 36. 42Ibid., pp. 38-39. 43Mary A. Holman, "A National Time-Budget for the Year 2000," Sociology and Social Research 46 (October 1961): 24. 44Staikov, "Time-Budgets and Technological Pro- gress," p. 479. 4SSzalai, The Use of Time, p. 10. 46Ibid., pp. 525-28. 47Brail and Chapin, "Activity Patterns of Urban Residents," p. 169. 48Chapin, Human Activity Patterns, pp. 43-49. 49Summarygof United States Time Use Survey, Survey Research Center, Institute for Social Research, University of Michigan, Ann Arbor, Michigan, 1966, p. 1. 50Ibid., p. 3. 51Ibid. 52Ibid. 53Ibid., pp. 9-10. 54Szalai, The Use of Time, pp. 562-66. 55Chapin, Human Activity Patterns, p. 70. 56Philip J. Stone, "Models of Everyday Time .Allocations," in Tge Use of Time, ed. A. Szalai (The Hague: Mouton, 1972): PP. 179-82. 149 57Ibid., pp. 183-88. 58Chapin, Human Activity Patterns, p. 11. 59George C. Hemmens, "Analysis and Simulation of Urban Activity Patterns," Socio-Economic Planning Sciences 4 (1970): 53. 6OChapin, Human Activity Patterns, p. 16. 61Janet Tomlinson, N. Bullock, P. Dickens, P. Steadman, and E. Taylor, "A Model of Students' Daily Activity Patterns," Environment and Planning 5 (1973): 231-32. 62Ibid., p. 265. CHAPTER V THE SIMULATION MODEL Microanalytic Approach The computer simulation model developed to con- duct the telecommunication experiments is called TIMMOD, named from its functional description as a time model. TIMMOD is a microanalytic model beginning with a single individual as the unit of analysis and building popu- lations one individual at a time. This is contrasted with a macroanalytic model, which would begin with a group of individuals as the unit of analysis. The essential logic underlying TIMMOD is to develop a series of probability distributions which describe a single individual's decision-making regarding the allocation of time from real world decision processes. In this case data describing the real world decision processes are derived from the 1965 United States Time Use Survey. An individual's day is then constructed by the computer by drawing random deviates from the various empirically derived probability distributions. Populations are created by merely repeating the process for subsequent 150 151 individuals. To experiment with the model it is only necessary to manipulate an appropriate probability dis- tribution and run the model. The fundamental operation of TIMMOD is borrowed from a micro-level activity system simulator designed by Brail.l Brail's model, while developed only to the extent of a generalized block diagram, demonstrates the logic of working with an individual decision-maker and developing aggregate measures by running a number of individuals through the same process as the first. Brail's model also points to a very important distinction between activity selection and activity duration. Time, according to Hemmens, enters into the structure of daily activity patterns in two ways. First is the duration of the activity, and second is the time of occurrence of the activity.2 The duration of an activity is simply the amount of time given to the activity and is of course of basic building block of any time allocation model. The time of day is equally important in that it affects directly the selection and sequencing of activities, as well as their duration. At a minimum, then, it is necessary to include activity selection probabilities that vary with the time of day in the allocation model. Ideally the activity duration probabilities would vary with the time of day as well, but this is not possible due to the limited sample 152 sizes of available time-budget surveys. This problem will be discussed in more detail later. The basic building blocks of a time allocation model are probability distributions which describe both the selection and duration of human activities. The logic of the Brail model is diagrammed in Figure 3. The model begins by determining a waking time for one individual from a continuous probability distribution. The individual then moves to an activity choice dis- tribution which is divided into two subsets of in-home and out-of—home activities. Since the number of activi- ties which can be chosen is finite, Brail suggests the use of a discrete probability distribution for activity choice. If the activity chosen is out-of—home, then a location must be determined by sampling another dis- crete probability distribution which would account for the present location of the individual as well as the location of the residence. The travel time necessary is determined by sampling a continuous probability dis- tribution describing the apprOpriate travel time. In Brail's model activity choices are considered as discrete events and elapsed time as a continuous event. After the activity has been selected, whether it be in-home or out-of—home, the duration of the activity is determined from a continuous probability distribution describing that particular activity. 153 on» ) flaking Tine Activity Choice Out Where? location Choice In Travel i": 1 Time Yes Activity No Retire? Duration Call Sun Accounting Retiring Times Subroutine Potential Fig. 3. Brail's Activity System Simulation Model SOURCE: Richard K. Brail, "Activity System Investi- gations: Strategy for Model Design" (Ph.D. dissertation, University of North Carolina, 1969). 154 If any travel time was involved, it is added to the activity duration time. Brail's flow diagram now passes control to an accounting subroutine which keeps track of the current time of day, the activity choice and activity location as the day progresses. Upon return from this subroutine, the retiring potential for the individual is determined from a continuous probability distribution. This is followed by a retiring decision from a discrete probability distri- bution. If the individual is not retiring, then another activity is selected and the process is repeated. If the individual is retiring, then a "new" individual is started with the sampling of another waking time. The process is continued until the desired number of hypo- thetical individuals have been cycled through the model. Brail suggests two general kinds of experimen- tation that would be appropriate with the micro-level activity system simulator. First is the study of changes in activity choice and structure. This includes both the substitution of one activity for another as well as the temporal patterning of the sequence of activity throughout the day. Second is the study of the changes in location choice.3 It is the first kind of experimentation that is the primary focus of TIMMOD, that is the substitution of one activity for another. Examples of such experimentation in this BCN technology 155 forecasting setting would include the substitution of telecommunication for transportation and the substi- tution of interactive television services for various face-to-face services such as shopping and marketing. The consideration of location is of much less importance here since electronic telecommunication is virtually independent of out-of—home distance. Brail cites two basic problems with the micro- level activity system simulator that he has proposed. First is the very difficult problem of simulation model verification. Computer simulation models are by their very nature extremely complex models requiring verifi- cation of their ability to imitate real world systems before any confidence can be attached to their output. The verification problem not only involves the compre- hension of the simulation model itself, but also basic issues in the philosophy of science. The problem of model verification will be discussed in more detail later in this chapter in the context of TIMMOD. The second problem described by Brail is in the determination of the probability distributions required by the model. It would be difficult enough to develop the discrete probability distributions required for the activity and location choice decisions, according to Brail, without considering the need to alter the probabilities as they may change throughout the day.4 156 Determining the probability distributions would first require a large body of empirical data able to support the cross-classification necessary to provide statisti- cally stable estimates of the probability of activity choice by hour of the day. The problem is further complicated when stable estimates of the duration of activities are also required by hour of day, as well as by location such as in-home and out-of—home. Unfor- tunately, none of the time-budget surveys described in Chapter IV are sufficient to develOp the probability distributions required by Brail's model. They are simply too small. Even if sufficient data were available, deter- mining probability distributions would still represent a major problem. To begin with, it would have to be decided to use an empirically derived distribution taken directly from the data or a theoretical probability dis- tribution, such as the normal distribution. Generally speaking, a theoretical probability distribution would be preferred to an empirically derived distribution because of computational convenience and simplicity. Assuming that a known theoretical probability distri- bution is adequate, then comes the problem of selecting the appr0priate probability distribution. For example, if a discrete distribution were appropriate a Bernoulli or Poisson distribution could be selected, or if a con- tinuous probability distribution were appropriate a 157 uniform or beta distribution could be selected. There are literally dozens of families of theoretical proba- bility distributions to choose from. Assuming that the appropriate theoretical probability distributions are selected, then comes the problem of establishing the various distribution parameters. No doubt the major reason why Brail's model has not progressed beyond the block diagram stage is the barrier raised by the probability distribution problem. Even in TIMMOD the probability distribution problem has been only partially solved because of the limitations imposed by a less than adequate empirical base. Several compromises are necessary to develop a working computer model within the constraints of available data. A functional description of TIMMOD indicating some of the problems will precede a continued discussion of theoreti— cal probability distributions. A Description of TIMMOD The computer simulation model, TIMMOD, is a working computer program designed specifically for the purpose of conducting the experiments designed to evaluate the potential impact of BCN technology. TIMMOD grew out of Brail's micro-level activity system simulator, and is itself a micro-level activity system simulator. TIMMOD is designed to simulate the activity of an average adult during an average day living in the 158 United States. The average adult is a composite of working men, working women, and housewives. The average day is a composite of weekdays, Saturdays and Sundays. TIMMOD does not include unemployed adult men or children, and also does not represent special holidays such as Christmas. TIMMOD is intended to represent the average day spent by the average American adult. Two significant departures from Brail's micro- level activity system simulator are necessary to develop TIMMOD into a working computer model because of the constraints imposed by a limited empirical base. The first of these is the assumption that activity duration does not change with the hour of the day. This assumption means that the activity event of eating or watching tele- vision will be assumed to be the same duration during the morning hours, as during the afternoon hours, as during the evening hours. The probability of occurrence does change with the hour of the day; only the duration of the activity event remains constant. This means that the length of time eating breakfast during the eight o'clock in the morning hour is the same as eating dinner in the six o'clock in the evening hour, but the proba- bility of eating breakfast at eight o'clock is not the same as eating dinner at six o'clock. The reason for this simplifying assumption is the lack of available data to generate activity duration statistics by hour 159 of the day, requiring activity duration statistics to be compiled and used summed across the entire day. A second departure is the lack of any logical connection between travel and in-home and out-of—home activities. TIMMOD considers travel to be the same as any other activity, and does not require the selection of an out-of—home activity requiring travel to be selected first. Again the reason for this change is the lack of available data describing the linkage between activity choice and physical location. It is simply not possible to adequately describe physical location using existing activity choice categories. While these simplifications may lead to dis- tortions in the output of the model, they do not negate the model's utility. Remembering that the simplifications are dictated by empirical necessity, it is possible to sufficiently limit experimental manipulation of the model so as to minimize the effect of any distortion. TIMMOD is not able to determine changes in the temporal sequenc- ing of activities, and must be limited to aggregate activity times for the entire day. For example, eating as an activity is considered only as a total for the entire day as a sum of breakfast, lunch, dinner, and any snacks. This allows any differences in activity duration at different times of the day to average together and approach the overall activity duration 160 used in the model. For example, if breakfast is always a little too long and dinner always a little too short, the total eating time will be the same. This still, however, does not rule out potential distortion in the activity selection process. An activity being too long or too short can change the probability of another activity being selected because the activity selection probabilities do change with the hour of the day. This effect is probably minimal for two reasons. First is that activity selection probabilities do not dramatically change from hour to hour except in only a few cases, and TIMMOD output would be interpreted only for rela- tively large numbers of hypothetical individuals causing the effect to average away across individuals. TIMMOD is also unable to determine any changes in the physical location of various factors in the simu- lated environment. This limitation on the application of TIMMOD is not important to this experiment because of the focus on telecommunication rather than changes in physical distance. Travel time is considered in the same way as any other activity making it an ideal situation for experimental substitution. However, there is a potential for distortion because travel and activity requiring travel are not logically related. It is possible to distort the impact on other activity caused by a change in travel time. The only solution 161 to this problem is care in the selection of experimental treatments and the consideration of only relatively large numbers of hypothetical individuals, assuming that the effect may average away. It should also be pointed out that whatever dis- tortions may exist in TIMMOD, they are held constant in both control and experimental treatment conditions. Assuming the worst case with the maximum distortion caused by the simplifying assumptions, then the experi- mental results would be interpreted as coming from a hypothetical world where all like activity events have the same duration throughout the day and travel is con- sidered an independent activity in the same way as all other activity. No doubt this is not the case, but some caution should be employed in interpreting TIMMOD output knowing the underlying assumptions. A number of computer simulation languages are available to write the computer program necessary to make a model like TIMMOD a working model. Several such special languages are available for general use at Michigan State University on the CDC 6500. Among the available simulation languages are SIMULA-67 which is based on ALGOL-60; SPURT (Simulation Package for Uni- versity Research and Teaching) which is a package of FORTRAN subroutines providing stochastic generators, list processing capabilities, and scheduling functions 162 for controlling discrete time simulations; and GASP II (General Activity Simulation Program) which is a col- lection of FORTRAN subroutines designed for discrete event simulation.5 TIMMOD while not written using all of the GASP II subroutines is written in Control Data Extended FORTRAN using the GASP II random variable generators and the GASP II approach.6 GASP II, developed by Pritsker and Kiviat, is a set of FORTRAN programs and subroutines organized to provide seven specific functional capabilities required by every simulation: 1. Event control 2. Information storage and retrieval 3. System state initialization 4. System performance data collection 5. Program monitoring and event reporting 6. Statistical computations and report generation 7. Random variable generation7 To enhance program efficiency, the TIMMOD main program accomplishes the GASP II tasks of event control, system performance data collection, and program monitoring and event reporting. Since the information requirements of TIMMOD are relatively small because information is only stored in aggregate form, the information function is accomplished through the liberal use of common storage. A system state initialization subroutine called Setup and a statistical computation and report generation subroutine called Report have also been written to 163 support the main program. Random numbers are provided through the use of GASP II routines used as external FORTRAN functions. The TIMMOD program accommodates thirty—two dif- ferent activity categories spread across twenty one-hour time periods beginning at five o'clock in the morning and extending past midnight. One TIMMOD run for a sample of 1,000 hypothetical individuals would require the generation of nearly 50,000 random numbers. The TIMMOD main program as shown in Figure 4 begins with a call to the Setup subroutine which initializes all of the system variables. The operation of the Setup subroutine will be described later. TIMMOD's next task is to sample a random number for the waking time and then the retiring time for the first individual or cycle. Another random number is drawn to determine whether or not this individual will work at any time during the day. If the individual does not work, then the work-related activities are eliminated from the activity selection matrix, if the individual does work then the work-related activities are included along with all other activities. The waking time is converted to an integer value and the activity selection matrix is entered at that point. For example if the waking time were 6.75, the activity selection matrix would be entered at 6 which represents the six o'clock Cm D ‘ Call Setup 1 Determine Hake Time Determine Retire Time No Yes 7 Eliminate work Activity l 1‘“ 1154 Determine Activity Duration Increment Time l Store Activity Duration J, Increment Activity . Counter Activity “‘3 Before Balance Activity Duration Increment Activity Counter Activity Ye? Before 7 Increment 1 Participation ' Counter Select Increment End Participation Activity Count er 1 J4 Fig. 4. TIMMOD Main Program Flow Diagram 165 hour. It should also be noted that the time and clock are maintained in hours and a decimal fraction to facili- tate arithmetic within the program. The time of 6.75 would obviously be equal to 6:45. An activity is selected randomly according to the appropriate hour. An activity duration is then determined from the duration probability distributions. If the current activity duration added to the current time, in this case the waking time, does not exceed the retiring time then the activity time is permanently added to the current time, thus incrementing the clock to account for the current activity. It should be noted that the clock is kept in decimal or floating-point form, and not in the integer form. The integer conversion is made each time it is necessary to enter the activity selection matrix. After the clock has been incremented, the activity duration is added to other of the same activity for this individual and stored. If this is the first occurrence of this particular activity for this individual then the participation counter is set to 1, if not this step is omitted. The current time is then converted to an integer value and the activity selection table is entered and a new activity is selected. This is followed by another activity duration selection, and through the rest of the process until the current time added to the current activity duration exceeds the retiring time. 166 When the retiring time is exceeded, the current activity duration is forced equal to the difference between the current time and the retiring time balancing out the day. This procedure is different from the one suggested by Brail that requires drawing two random numbers after each activity selection to determine retiring time. The simple balancing procedure used by TIMMOD is computationally simpler and faster, although it may lead to some distortion in the duration of the last activity before retiring for the night. This, however, should make little difference because the activity duration times are held constant throughout the day, and TIMMOD runs are for large numbers of hypothetical individuals which should average out the effect. After the activity duration has been balanced, then the activity counter is incremented for the last activity category. If this is the first occurrence of the particular activity, then the participation counter is set to 1, just as before. If it is not the first occurrence of the activity, then this step is omitted. Since it is the last activity for the first individual, a decision is made based on an input parameter whether or not to continue with a new indi- vidual or cycle, or end the program. In both cases the generated data are stored in an array to be passed to 167 the report generating subroutine. TIMMOD also has the capability to select one additional activity category apart from the overall tabulation and generate statistics for participants in that category only, such as indi— viduals who go shopping or watch television. If the last cycle or individual has been reached, then a call is issued to the Report subroutine which produces the output reports. If the last cycle has not been reached, then TIMMOD returns to the beginning and selects a new waking and retiring time and reinitializes all of the individual arrays. The current time is reset to the waking time. This process is repeated until the number of cycles is exhausted. Figure 5 shows the flow diagram of the Setup subroutine which is called at the start of the TIMMOD run to initialize all of the system variables. Setup begins by dimensioning and initializing all of the system variables and arrays to zero. Setup then reads the run parameter card. The run parameter card spe- cifies the random number seed, the number of parameters included in the activity duration parameter list, the number of parameters included in the activity selection parameter list, the number of cycles to be run, and the activity category to be separately tabulated. The run parameters are then set to the appropriate system variables. ( 3 Initialize Variables and Arrays Set Activity Duration Parameters Read Run Parameters ) Read Selection Parameters Calculate Cumulative Select'Matrix Set Run Parameters _1__ Set Activity Selection Parameters Read Duration Parameters Convert to Hours Print Setup Parameters Fig. 5. TIMMOD Setup Subroutine Flow Diagram 169 The Setup subroutine then reads the activity duration parameters according to the run parameter card. The activity duration parameters include the theoretical probability distribution parameters, minimum and maximum values, and a weighting factor for each of the activity categories indicated by the run card. Since the activity duration parameters are entered into the TIMMOD program in minutes, it is necessary to convert all of the parameters in minutes to hours to facilitate compu- tation in the running of the TIMMOD main program. The appropriate arrays are then set with the activity dur- ation parameters. The last cards read by the Setup subroutine are the activity selection parameters. These cards, again determined by the run parameter card contain the average frequency of occurrence of a particular activity by hour of the day. As will be described later, it is necessary to convert these frequencies to cumulative percentages by hour of the day. Setup performs all of the necessary calculations and then sets the appro- priate array equal to the cumulative values. Setup then prints a listing of all of the input parameters, including the cumulative activity selection matrix, and returns control to the TIMMOD main program. Figure 6 shows the flow diagram for the Report subroutine. Report is called by the TIMMOD main program 170 Convert ET . Time to Start Minutes Calculate Participation Percentages (/"“ J~ J1 Calculate Aggregate Average Time Activity Spent Categories 4 Calculate Time Spent Variance *1 Calculate Time Spent per Participant Return Fig. 6. TIMMOD Report Subroutine Flow Diagram 171 after the last cycle has been completed. Report begins by calculating the participation percentages from the total participation arrays passed to the subroutine from the main program. Then Report calculates the average time spent and the variance for the averages time spent for each activity category. The time spent per partici- pant is then calculated by dividing the participation percentage into the average time spent per participant for each activity category. All time estimates are converted from hours back into minutes. The TIMMOD report is then printed for all participants and for participants in the single category specified on the run parameter card. The TIMMOD report includes the number of times each activity occurred for the entire group of participants, the average time spent and variance, the percentage participating, and the average time spent per participant. Samples of the TIMMOD report are included later in this chapter. To simplify comparison with treatment and control conditions, a summary report is also included. This report is simply an aggregation of the activity cate- gories shown in the TIMMOD report and a calculation of the average time spent and variance for the new cate- gories. The TIMMOD summary is then printed. Report transfers control back to the main program which then ends the run. The FORTRAN source code for the TIMMOD 172 main program, the Setup and Report subroutines, and the external random number functions is in the Appendix. The functional description of TIMMOD at this point has ignored the critical problem of probability distributions. Each of the random numbers used in the operation of the model must come from either an empiri- cally derived or theoretical probability distribution. The selection and use of probability distributions is crucial to the operation of TIMMOD, requiring the development of a rationale for the selection. What follows is a brief discussion of the common probability distributions that are available to TIMMOD. Theoretical ProbabilitygDistributions TIMMOD is a stochastic simulation which requires the construction of a probabilistic model of the time allocation process. This kind of simulation is often termed "Monte Carlo" simulation and was first employed by mathematicians to find solutions to deterministic problems whose answer is not easily obtainable using standard numerical methods. The classic examples of the application of stochastic simulation to solve such problems are in the evaluation of multiple integrals, solutions to high-order difference equations, complex queueing problems, and job-shop scheduling problems. In general, Naylor, Balintfy, Burdick, and Chu describe stochastic simulation as involving the 173 replacement of an actual statistical universe of ele- ments by its theoretical counterpart, like the normal probability distribution, and then sampling by means of some type of random number generator from the theoreti— cal probability distribution.8 TIMMOD requires the sampling of random numbers from probability distributions representing the occurrence of activities throughout the day and the duration of each activity category. The problem is then to develop the appropriate probability distributions. Probability is normally considered as the long- run relative frequency of occurrence for some event or experiment, such as the tossing of a fair coin. A formal mathematical definition of probability is offered by Hughes and Grawoig: If an outcome occurs f times out of n trials, its relative frequency is f/n; the value which is approached by f/n when n becomes infinite is called the limit of the relative frequency. The probability of an outcome 0. is defined as the limit of its relative frequency; that is: P(Oi) = lim f/n9 n + m The relative frequency of the occurrence of an event is then the ratio of the number of times the event occurred to the total number of events. If all possible events are grouped together, then a distribution of relative frequencies is obtained. It is then a simple matter to convert a distribution of relative frequencies into a distribution of probabilities. 174 Probability distributions applied to computer simulation are of two general types, theoretical and empirically derived. Obviously the shape of a proba- bility distribution can vary dramatically with the set of events or phenomenon it is supposed to represent. Theoretical probability distributions can be represented by a mathematical function or rule which allows the generation of the entire probability distribution. Empirical probability distributions cannot be adequately described by a mathematical function, and require enum- eration of each event to generate the entire probability distribution. An empirical probability distribution is somewhat more awkward than a theoretical probability distribution from a computational vieWpoint, especially if a large number of events is involved. An empirical probability distribution also assumes that the data are available to generate the distribution, which is not always the case. I Empirical probability distributions have to be ruled out for TIMMOD because of the data problem. A sample of just over 1,200 individuals simply cannot provide stable estimates of frequency distributions for thirty different activity categories across twenty hours of the day. DeveloPment of the required proba- bility distributions using the available data necessi- tates the use of theoretical probability distributions. 175 Naylor et a1. recommend always attempting to use theoretical probability distributions first, with empirical distributions used only as a last resort.10 Theoretical probability distributions divide into two major categories, discrete and continuous. Discrete probability distributions describe events or variables that can take only discrete, non-negative integer values. Discrete probability distributions generally are used to describe counting processes which can be either finite or infinite, but are limited to whole numbers. Tsokos defines a discrete one-dimensional random variable as follows: Let X be a random variable. If the number of ele- ments in the range space, RX, is finite or countably infinite, then X is called a one-dimensional dis— crete random variable. A continuous random variable takes on uncountably infinite values, such as distance and time, and is limited only by the precision of the measuring instru- ment. Continuous random variables are considered in terms of intervals rather than discrete points. Tsokos defines a continuous one-dimensional random variable as follows: Let X be a random variable. If the range space, Rx' is an interval or the union of two or more non- overlapping intervals, then X is called a one- dimensional continuous random variable.12 176 To avoid the problem of working with an interval in performing calculations, continuous distributions are generally assigned values in a manner similar to the discrete case. Theoretical probability distributions can be described in a number of ways. To begin with, proba- bility distributions can be described in terms of their graphical representation and shape. Probability distri- butions can also be described algebraically using a mathematical probability function. Probability functions are rules for assigning the selection chances to the outcomes of a particular experiment. The probability function describes the mathematical behavior of a theo- retical probability distribution. Probability distributions are normally described in terms of f(Xi) which represents the distribution of a random variable X. The distribution of f(xi) is also referred to as the mass or frequency density function. The general probability density function for a one- dimensional discrete random variable is expressed as: The general probability density function for a one- dimensional continuous random variable is expressed as: f_: f(x)dx = l 177 Probability density functions of these general forms then are used to describe the various theoretical proba- bility distributions to be considered in this research. In addition to the probability density function it is also useful to consider the long-run average result or expected value of the various probability distributions. Expected value, expressed as E(X), of the random variable X is synonymous with central tendency and is the single most representative value of the distribution. Expected value, however, really does not adequately describe a distribution because it gives no indication of the dis— persion or range of values in the distribution around the expected value. Statistically, the variance is represented by the sum of the squared deviations around the mean divided by the total number of observations. The variance of random variable X is expressed as: _ 2(x - X)?‘ _ n V(X) Variance can also be expressed in the form of a standard deviation, which is represented as the square root of the variance. Theoretical probability distributions now can be characterized in a number of ways, distributional shape, probability density function along with any associated parameters, expected value, and variance. 178 To simplify the discussion of probability distributions, then, the above characteristics along with some appli- cations for some common "families" of theoretical probability distributions will be only considered. Discussing probability distributions in terms of families that share certain characteristics is suggested by Zehna, and the families that follow are his catalog of common families.13 Discrete Probability Distributions Discrete probability distribution families are those that describe various counting processes. The following five families are the most common. Bernoulli Family The Bernoulli or Point Binomial family of proba- bility distributions is named after the Swiss mathema- tician Jacques Bernoulli and is the probability law for coin—tossing experiments. It can be applied to any dichotomous random experiment, such as the case of a piece of equipment which either succeeds or fails. The probability density function for the Bernoulli family is as follows: f(x;p) = pqu-x The parameters include p, which is the probability of success for a given event, and must take a value 179 between 0 and l. The other parameter, q, is merely the probability of failure and is derived directly from p, q = l - p. The function generates values for X in the range of 0 to l. The expected value and variance for the Bernoulli family are expressed as follows: E(X) H *U V(X) Pq Binomial Family The binomial family is an algebraic generali- zation of the Bernoulli family adding the positive integer n to represent the number of trials. The appli- cation of the binomial family is similar to the Bernoulli family. The probability density function for the binomial family is as follows: f(x;n,p) := (:2) pan-X The parameters are the same as those for the Bernoulli family. The function generates values for X in the range of O to n. The expected value and variance for the binomial family are expressed as follows: ll E(X) np V(X) an 180 It should also be noted that for very large values of n, the binomial approaches the continuous normal distri- bution. Geometric Family The geometric family describes the distribution of the number of trials needed to achieve success. An example of the application of the distribution might be estimating the number of cycles a machine might operate before a failure. The probability density function for the geometric family is as follows: f(x;p) = pqx-l The parameters are again the same as the Bernoulli family. The function generates values for X in the range of l to infinity. The expected value and variance for the geometric family are expressed as follows: _ l E(X) — p V(X) = 95 p Negative Binomial Family The negative binomial family describes the number of repetitions necessary to achieve r successes. An application of the distribution is in inventory management. The total demand for an item of a given 181 type is normally assumed to be a random phenomenon. Especially in cases where the average demand is large and there is little past history the total number of units demanded is often assumed to be distributed according to the negative binomial distribution.14 The probability density function for the negative binomial family is as follows: x + r - l f(X;r,p) = ( X ) prqx The parameters remain the same, except for the addition of r which indicates the number of successes. The function generates values in the range from 0 to infinity. The expected value and variance for the geo- metric family are expressed as follows: = £9 E(X) p V(X) = £3 p Poisson Family The Poisson family, named after the French mathematician S. Poisson, is a limiting form of the binomial family. The parameter of the Poisson family is A, which is the mean number of occurrences of an event per unit of time over a given number of trials. The distribution assumes different shapes depending on 182 the value of A. When A is less than 1, the distribution is highly skewed to the right, and becomes more symmetri- cal as 1 increases. The Poisson family describes situ— ations when the concern is with the number of times an event occurs over some time interval. A large number of applications have been suggested for the Poisson family. Some examples are listed by Tsokos: l. the frequency of certain peaks per minute at a telephone switchboard; 2. the number of misprints per page in a dictionary; 3. the number of traffic accidents which occur per day on a certain turnpike; 4. the number of alpha-particles emitted per hour by a radioactive source; 5. the number of babies born with heart defects in a large city during a one-year period; 6. the number of no-hitters pitched by a Hall of Famer during his baseball career; 7. the number of live viruses remaining after the production process of a certain vaccine. The probability density function for the Poisson family is as follows: -1Ax foul) = e x, The parameter A is defined earlier. The function gen- erates values for X in the range of O to infinity. The expected value and variance for the Poisson family is shown below: II >2 E(X) V(X) 183 While other discrete probability distributions exist, such as the hypergeometric distribution, the families mentioned here represent the more commonly used distributions. Attention can now be turned to con- tinuous probability distributions. Continuous Probability Distributions Five major continuous probability distribution families will be discussed in the same manner as the discrete probability distributions. Other continuous probability distributions will receive only brief mention. Uniform Family The simplest continuous probability distribution family is the uniform family. The uniform distribution applies in situations where all events are equally likely to occur. As a model for random experiments, Zehna describes the uniform family as being suitable for bounded random variables whose essential range coincides with the interval (a, B).16 A uniform dis- tribution is shown graphically in Figure 7. Its dis- tributional shape is simply the representation of a horizontal line inside the range of its parameters. The probability density function for the uniform family is expressed as follows: f(x;a.B) = 184 ,HH ,....*“"1———" n l—--—-—- Fig. 7. The Uniform Distribution The parameters a and 8 set the lower and upper boundary for the random variable X. The expected value and variance for the uniform family are expressed as follows: _ a + B E(X) —- 2 2 (a - B) V(X) 12 Exponential Family The exponential family provides density functions for non-negative random numbers. A classic application of the exponential is estimating the time to failure of a machine. Like the Poisson family, the exponential is well suited to describe the occurrence of an event across time intervals. According to Naylor et al., if the probability that an event will occur in a small time interval is small, and if the occurrence of the event is statistically independent of other events, then 185 the time interval between the occurrence of events is exponentially distributed.17 An exponential distribution is graphically shown in Figure 8. The l is the parameter for the distribution like the Poisson distribution. Fig. 8. The Exponential Distribution The probability density function for the exponential family is expressed as follows: f(x;1) = le-AX The l parameter must be greater than 0. The function generates the random variable X in the range 0 to infinity. The expected value and variance for the exponential distribution are expressed as follows: >Jll—I E(X) V(X) .__1_ x2 Gamma Family The gamma family is a more general family of distributions for non-negative random variables. The gamma distribution has two parameters, a which is the 186 number of successes per interval or unit space, and B which is the reciprocal of the average number of successes per interval (%). The gamma distribution is related to both the Poisson and exponential distributions. The exponential is a special case of the gamma distri- bution when a = 1. As a increases, the distribution becomes less skewed until it approaches the normal distri- bution. One of the most powerful properties of the gamma family is the ability to change shape from an extremely skewed exponential distribution to a very near normal distribution by changing only the a parameter. As Zehna points out, the gamma family is so extensive that it is a fairly safe assumption to make as a model for an experi— ment described by almost any nonnegative random vari- able."18 Letting a = n/2, where n is a positive integer, and B = 2, the gamma distribution becomes a chi-square distribution with n representing degrees of freedom. Figure 9 shows the affect on the shape of the distribution by changes in the parameter a. .75‘ .501 .25‘ 187 The probability density function for the gamma distri- bution is expressed as follows: f(x-a B) = Xoc-le-x/B bar(a) The a and 8 parameters have been described before, and they both must be greater than 0. The F notation indi- cates a one-parameter integral called the gamma function as shown below: r(p) = I: xp‘le’xdx In this function, p must be greater than 0. The gamma probability density function generates a random variable X in the range from O to infinity. The expected value and variance of the gamma family are expressed as follows: E(X) a8 V(X) = a82 Beta Family The beta family makes possible structuring a model for a random variable in the range 0 to 1. It is a two parameter distribution like the gamma family, and is rich enough to take almost any shape. Also like the gamma, the beta family is related to a function called the beta function. However, a beta function can be 188 expressed in terms of gamma functions. The beta family flexibility is illustrated in Figure 10 which shows dif- ferences in the value of a with a fixed 8. The beta family probability density function is expressed as follows: F(a + B) a-l F(a)P(B) f(x;a,B) = (l - X)B-l Fig. 10. The Beta Distribution The affiliation of the beta family with the gamma family is apparent. The a and 8 parameters must be greater than 0. The function returns a random variable X in the range of O to l. The expected value and variance of the beta family is expressed as follows: C! E(X) = a + B V(X) = 20‘5 (a + B) (a + B + l) 189 Normal Family The last continuous probability distribution family is probably the most important. Many continuous variables such as height, weight, and intelligence quotient are normally distributed. The normal family is the most important probability model is statistical analysis, and no doubt the most familiar. It is shown graphically in Figure 11 with three different values for the standard deviation, or c. The mean u is held constant at O. The probability density function for the normal family is expressed as follows: 1 - 2 e——§iX U) f(X;UI02) = W626 Fig. 11. The Normal Distribution 190 The normal of Gaussian family is a two-parameter family, with the familiar mean, u, and variance, 02. The expected value and variance are obvious: E(X) = u V(X) ll Q A large number of other continuous probability distributions could also be used. Among these are the Cauchy distribution, the Laplace distribution, the log- normal distribution, the Weibull distribution, the Rayleigh distribution, Maxwell's distribution, extreme- value distribution, arc sine distribution, and Pareto's distribution. None of these have application to TIMMOD, so the discussion will be terminated. Additional dis- cussion and description can be found in Tsokos.19 Applications to TIMMOD There are two general approaches to the gener- ation of random numbers in TIMMOD. The first is the use of a continuous uniform distribution to determine the activity choice. The second is the use of a gamma distribution to determine the duration of the activity. The activity selection process is very similar to the media selection process used by the Simulmatics Model described in Chapter III. The activity selection frequencies are arranged in a two-dimensional matrix, activity by time of day. The activity selection 191 frequencies are converted to cumulative percentages, beginning with the first hour of the day through the last for each activity category. A random number is then sampled from a uniform distribution bounded by 0.0 and 100.0. This number is then skipped through the cumulative percentages for the appropriate hour of the day as determined by the logic of the computer program. The activity selected is the activity category where the random number lies. While the essential nature of this selection process is discrete, it is computationally convenient to use the continuous probability distribution to compare percentages. A similar procedure is used to determine work activity. The determination of activity duration directly involves sampling the continuous variable time. Activity durations are represented in a wide range of distribu- tional shapes, from extremely skewed J-shaped distribu- tions to a normal distribution. The range of activity duration can extend from 0 to 1,440 minutes or 24 hours. The appropriate theoretical distribution to describe activity duration is the gamma family. A gamma distribu- tion can describe non-negative random processes from extremely skewed exponentially shaped distributions through a symmetrical approximately normal distribution. The versatility of the gamma family makes its appli- cation popular. Basic, for example, uses the gamma 192 distribution to approximate demand for products in the metal service industry and then exploits the versatility of the gamma distribution family to serve as a guide for the management of inventory within the industry.20 To use a gamma distribution all that is required are the two parameters, a and B, which have been described previously. These parameters, with some effort, can be estimated from directly from an empirical data base. The calculation of gamma random variates, however, is reasonably complex and requires some discussion. Computer Generated Random Numbers There are essentially four methods of generating random numbers. The methods include manual methods, such as coin tossing and card shuffling, random number tables, the use of analog computers, and the use of digital computers. Since TIMMOD is programmed for a digital computer, only modes for generating random numbers using a digital computer will be considered. Three alternative modes for generating random numbers are discussed by Naylor et a1.21 External pro- vision of random numbers is possible by storing a random number table on a peripheral device, such as a magnetic tape or disk unit. This mode, however, is quite slow because of the time required to read the device. A second mode is the internal generation by a random physical process such as the use of the noise generated 193 by some electronic process. The major problem with this mode is inability to reproduce the process so that calcu- lations can be checked. The third, and most widely used mode, is the internal generation of random numbers through the use of a recurrence relation. The recurrence relation involves a mathematical transformation of a group of arbitrarily chosen numbers. Normally the starting number for the sequence is referred to as the seed. The selection procedure is a determi- nistic process. The same random number seed will gen- erate the same sequence of numbers, although the next number in a sequence cannot be predicted. Because com- puter generated random numbers are drawn using a deter- ministic process, they are commonly referred to as pseudo- random numbers. However, the pseudorandom process must generate random numbers which provide sequences which must meet several stringent requirements. Nayor et a1. enumerate five criteria for an acceptable method of generating random numbers including, uniform distri- bution, statistical independence, reproducibility, high speed, and low core memory requirements.22 Random numbers for TIMMOD are selected using a congruential method for TIMMOD represented by the CDC FORTRAN function RANF.23 The functions for generating deviates of the uniform and gamma distributions are borrowed from 194 Pritsker and Kiviat.24 The functions themselves are shown with the FORTRAN source code in the Appendix. Both functions employ the inverse-transformation method suggested by Naylor et al. to calculate the random deviates.25 The uniform distribution is handled in a straight- forward manner and is quite simple. The gamma distri- bution, however, is different and requires additional discussion. As Naylor et a1. point out, the cumulative distribution function for a gamma distribution cannot be formulated explicitly, requiring an alternate method.26 Generally the approach is to limit the consideration of the gamma distribution to an Erlang distribution, which is a gamma distribution with an integer parameter. The a parameter then is limited only to integer values. Mathematically the Erlang distribution is a convolution of a exponential distributions. When a = 1, the Erlang obviously becomes an exponential, as 0 increases in integer increments the Erlang approaches the normal distribution. A convenient computational formula is provided by Naylor et al., which is expressed below and used in the TIMMOD model:27 195 The random variate is x, a and B are the gamma para- meters, and r is the pseudorandom number provided by RANF. The TIMMOD functions are appropriately named UNIFRM and ERLNG. Preparation of TIMMOD Input Data The input data for TIMMOD is estimated from the 1965 United States Time Use Survey described in Chapter IV. Data from this study are reported in two sources, Summary of United States Time Use Survey28 and as part of the multinational comparative study described in The Use of Tipg.29 The most useful tabulations for developing input for TIMMOD are found in the Statistical Appendix of the latter source. Unfortunately these tabulations do not directly provide the data necessary for TIMMOD, namely mean and variance estimates for the duration of each activity event for the activity duration statistics and the frequency of occurrence for each activity by hour of the day for the activity selection matrix. All that is provided for activity duration statistics is the total time spent engaging in the activity for the entire sample, the percentage participating, and the average number of occurrences for participants. No variance estimates or frequency distributions are provided, except for a very few aggregate categories. Activity selection statistics are available only for aggregate 196 categories, and not for the entire sample, only sub- groups. In order to estimate the input parameters for the model it is necessary to perform a number of calcu- lations on the available data. This transformation process will be described in detail. The problem of variance estimates was quickly solved by contacting the Institute for Social Research at the University of Michigan which provided them for a large number of activity categories for the total time spent in that activity. All other data come from the Statistical Appendix of The Use of Time. The next problem is to establish a consistent set of activity categories. The original United States Time Use Survey used ninety-six activity categories, but abridged this to a reduced thirty-seven activity categories for reporting the results. However, the reduced thirty-seven is still somewhat difficult to handle, and does not reflect all of the activities necessary to focus on marketing communication activity. The reduced thirty-seven activity categories require additional modification to arrive at the TIMMOD thirty- two activity categories. This last modification is shown in detail in Table 3. Marketing, which as an activity category refers to shopping for everyday nondurable goods, and shopping and errands are separated from other activity and 197 TABLE 3 COMPARISON OF THE MULTINATIONAL COMPARATIVE STUDY AND TIMMOD ACTIVITY CATEGORIES Multinational Comparative Time-Budget Project Reduced 37 Activity Categories TIMMOD 32 Activity Categories 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24 25 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. Main job 1. Second job 2. At work other 3. Travel to job 4. Total work Cooking 7. Home chores 8. Laundry 9. Marketing 5. Total Housework Garden, animal care 10. Errands, shopping 6. Other house 11. Other household obligations Child care Other child Total child care 12. Personal care 16. Eating 15. Sleep 32. Total personal needs Personal travel 13. Leisure travel 14. Total non-work travel Study Religion OrganiZation Study and participation 17. Radio 19. TV (home) TV (away) 18. Read paper 20. Read magazine 21. Read Books 22. Movies 23. Total mass media Social (home) Social (away) 24. Conversation 25. Active sports 26. Outdoors 27. Entertainment 28. Cultural events 29. Resting 30. Other Leisure 31. Total Leisure Main Job Second Job Work-Other Travel to Job Cooking Home Chores Laundry Marketing Pets and Garden Shopping-Errands Other House Child Care Personal Care Eating Night Sleep (less day sleep) Personal Travel Leisure Travel Study-Clubs Radio Television (23 and 24 combined) Newspapers Magazines Books Movies Social Activity (29 and 30 combined) Conversation Active Sports Outdoors Entertainment Cultural Events Resting-Naps (includes day sleep) Other Leisure SOURCE: A. Szalai, ed., The Use of Time (The Hague: Mouton, 1972). 198 highlighted. The child care category is aggregated, as is study, religion and organization. Sleep is divided into night sleep, with day sleep being combined with resting. Television at home and away from home are combined in one category as is social activity. The other categories remain the same. Given this set of thirty-two activity categories, the development of the necessary activity duration sta- tistics and the activity selection matrix will be described in sequence. Activitijuration Statistics The estimation of the mean and variance of a single activity event requires a rather complex sequence of calculations. Starting with the means and variances of the reduced thirty-seven categories requires some initial aggregation. The means and variances for the total time spent per sample member are easily combined using the following equations: Mt = M1 + M2 + --- + Mn 2 _w 2 2 ... 2 st — 51 + $2 + + 8 Each of these equations assumes that each of the cate- gories is mutually exclusive. The total mean, Mt’ is simply the sum of the means comprising it. The total 199 . 2 . . variance, S 18 the vector sum of the variances t' associated with the means in the total mean calculation. With the new structure of thirty-two activity categories established, the preparation of the activity duration mean and variance estimates begins. Data describing the average time spent for the total sample engaged in primary activities are shown in Table 4. The total sample is used to make TIMMOD representative of the overall adult population, not just a specific sub-population such as employed men, employed women, or housewives. A distinction between primary and secondary activities is made throughout the multi- national comparative study. A primary activity is the main activity recorded by the respondent, and the secondary or accompanying activity is then recorded in addition to the primary activity. The distinction between primary and secondary activity is left com- pletely to the respondent. For example, watching tele— vision and eating, or eating and watching television, define different primary and secondary activities in each case. TIMMOD assumes only primary activities, which no doubt understate certain activities such as radio listening and conversation. Conversation as an activity category in TIMMOD is as a primary activity only. 200 TABLE 4 PROVIDED ACTIVITY SELECTION STATISTICS Total Sample Time Spent Percentage Mean Per Activity Participating Participant Mean Standard Variance DeVlatlon 1. Main Job 225.0 236.1 55731.0 51.2 439.5 2. Second Job 4.9 37.0 1369.0 2.4 204.2 3. Work--Other 11.5 20.0 400.0 33.3 34.5 4. Travel to Job 24.9 36.0 1296.0 50.7 49.1 5. Marketing 14.0 31.0 961.0 34.4 40.7 6. Shopping--Errands 17.9 42.1 1775.0 30.5 58.7 7. Cooking 43.8 53.0 2809.0 60.6 72.3 8. Home Chores 58.4 63.3 4013.0 60.9 95.9 9. Laundry 25.5 53.2 2825.0 29.5 86.4 10. Pets and Garden 3.3 13.0 169.0 12.2 27.0 11. Other House 23.7 55.7 3097.0 37.7 62.9 12. Child Care 32.6 58.1 3370.0 37.6 86.7 13. Personal Travel 30.8 37.5 1410.0 63.5 48.5 14. Leisure Travel 19.3 32.2 1036.3 42.9 45.0 15. Eating 81.3 50.9 2589.0 98.5 82.5 16. Personal Care 68.5 45.6 2081.0 98.2 69.8 17. Study--Clubs 27.7 64.9 4214.0 20.3 136.5 18. Television 91.6 99.0 9801.0 69.7 131.4 19. Radio 3.6 18.0 324.0 8.2 43.9 20. Newspapers 23.8 32.0 1024.0 48.6 49.0 21. Magazines 6.4 21.0 441.0 11.3 56.6 22. Books 5.3 24.0 576.0 7.1 74.6 23. Movies 3.2 25.0 625.0 2.0 160.0 24. Social Activity 63.0 95.1 9040.0 54.6 115.4 25. Conversation 18.4 35.0 1225.0 41.2 44.7 26. Active Sports 5.5 27.0 729.0 5.7 96.5 27. Outdoors 2.4 18.6 345.0 3.1 77.4 28. Entertainment 5.4 30.5 928.0 3.7 145.9 29. Cultural Events 0.6 8.1 65.0 0.7 85.9 30. Resting-~Naps 19.4 45.4 2057.0 18.6 104.3 31. Other Leisure 19.5 44.9 2015.0 25.7 59.7 SOURCE: A. Szalai, ed., The Use of Time (The Hague: Mouton, 1972), and Summapy of United States Time Use Survey, Survey Research Center, Uni- versity of Michigan, Ann Arbor, Michigan, 1966. 201 The mean total time spent statistics are obtained directly from the statistical appendix of The Use of Time using the data reported for the study designated forty-four cities, U.S.A.3O The standard deviation and variance estimates supplied by the Insti- tute for Social Research are also shown in Table 4. The percentage of the total sample engaging in a par- ticular activity category is also taken directly from 31 The Use of Time. The percentage participating is also shown in Table 4. The average or mean time spent per participant is then easily calculated using the following simple equation: T = T .2 p P The average time Spent per participant, Tp, is calcu- lated by dividing the average time spent overall, To' by the percentage participating, P. This calculation is included in The Use of Time, but required recalcu- lation because of the revised TIMMOD categories.32 Estimating the variance for participants is somewhat more complex. Using the computational formula for calculating variance, the following estimation equation is derived: P(S§) + PT2 - T: O 2 p p 202 The variance for participants, 8:, is estimated from the proportion participating, P (expressed as a decimal fraction), the mean, To' and variance, 8:, for the time spent for the overall sample. The result of this calcu- lation is shown in Table 5. Now that the mean and variance for participants has been estimated, it is still necessary to reduce the estimates to the single activity level by taking into account the number of occurrences of a particular activity event per participant. These data are taken directly from the statistical appendix of The Use of Time and are shown in Table 5.33 The mean and variance for a single activity event are then estimated using the following equations: is Te = ]< 2 s 2.._P_ Se _' k The mean and variance for a single activity event are represented by T6 and 8: respectively. The mean and variance for all participants is the same as in previous equations, Tp and 8:. The number of occurrences is represented by k and obviously can never have a value of less than one. The results of this calculation are shown in the last two columns of Table 5. 203 TABLE 5 ESTIMATED ACTIVITY DURATION STATISTICS Occurrences . . Percentage Activity Activ1ty . . . Per . Part1c1pating Participant Duration Mean Variance Mean Variance 1. Main Job 439.5 14614.5 4.0 109.9 3653.6 2. Second Job 204.2 15833.3 1.5 136.1 10555.5 3. Work--Other 34.5 446.8 2.1 16.4 212.8 4. Travel to Job 49.1 1367.3 2.4 20.5 569.7 5. Marketing 40.7 1711.9 1.3 31.3 1316.8 6. Shopping--Errands 58.7 3426.9 1.5 39.1 2284.6 7. Cooking 72.3 2579.0 2.7 26.8 955.2 8. Home Chores 95.9 2993.0 3.2 30.0 935.3 9. Laundry 86.4 4309.2 2.3 37.6 1873.6 10. Pets and Garden 27.0 733.3 1.6 16.9 458.3 11. Other House 62.9 5758.5 1.7 37.0 3387.4 12. Child Care 86.7 4283.0 3.4 25.5 1259.7 13. Personal Travel 48.5 1362.8 3.1 15.6 439.6 14. LeiSure Travel 45.0 1259.2 2.5 18.0 503.7 15. Eating 82.5 2526.9 2.9 28.4 871.3 16. Personal Care 69.8 2032.2 3.0 23.3 677.4 17. Study--Clubs 136.5 5948.8 1.9 71.8 3130 9 18. Television 131.4 8824.9 1.9 69.2 4644 7 19. Radio 43.9 2189.6 1.3 33.8 1684.3 20. Newspapers 49.0 875.4 1.3 37.7 673.4 21. Magazines 56.6 1046.9 1.2 47.2 872.4 22. Books 74.6 2960.0 1.3 57.4 2276.9 23. Movies 160.0 6250.0 1.0 160.0 6250.0 24. Social Activity 115.4 10516.4 1.7 67.9 6186.1 25. Conversation 44.7 1797.6 1.7 26.3 1057.4 26. Active Sports 96.5 4062.5 1.0 96.5 4062.5 27. Outdoors 77.4 5100.0 1.2 64.5 4250.0 28. Entertainment 145.9 4414.3 1.1 132.6 4013.0 29. Cultural Events 85.7 2050.0 1.1 77.9 1863 6 30. Resting--Naps 104.3 2202.3 1.2 86.9 1835.3 31. Other Leisure 75.9 3565.2 1.5 50 6 2376.8 204 It is now possible to estimate the two para- meters for the Erlang probability distributions used to describe the characteristics of the activity duration distributions used by TIMMOD. The parameters a and B can be estimated using the expected value and variance equations for the gamma family described earlier in this chapter. The only modification required in using these equations is in holding the a parameter to an integer value. This is accomplished by rounding to the nearest whole number. The TIMMOD activity duration parameter estimates derived as a result of this procedure are shown in Table 6. Missing from Tables 4 through 6 has been category 32, night sleep. The reason for this omission is that night sleep is handled in a different manner in the logic of TIMMOD. As described before in this chapter, TIMMOD calculates both a waking and retiring time. The time for night sleep is then calculated from these times. The probability distributions for both waking and retiring times are Erlang distributions. The parameters from the distributions were estimated from the sleep category in the activity selection matrix which is described in the next section of this chapter. Both waking time and retiring time are highly skewed in the direction of sleep. Waking time is then generated from a highly skewed dis- tribution toward the early morning hours. Retiring time 205 TABLE 6 ESTIMATED TIMMOD ACTIVITY DURATION PARAMETERS . . . Erland Acthlty Duration Parameters Activity Mean Variance a B 1. Main Job 109.9 3653.6 19 5.78 2. Second Job 136.1 10555.5 15 9.07 3. Work--Other 16.4 212.8 5 3.28 4. Travel to Job 20.5 569.7 4 5.13 5. Marketing 31.3 1316.8 5 6.26 6. ShOpping--Errands 39.1 2284.6 5 7.82 7. Cooking 26.8 955.2 5 5.36 8. Home Chores 30.0 935.3 5 6.00 9. Laundry 37.6 1873.6 5 7.52 10. Pets and Garden 16.9 458.3 3 5.63 11. Other House 37.0 3387.4 4 9.25 12. Child Care 25.5 1259.7 4 6.38 13. Personal Travel 15.6 439.6 3 5.20 14. Leisure Travel 18.0 503.7 3 6.00 15. Eating 28.4 871.3 5 5.68 16. Personal Care 23.3 677.4 4 5.83 17. Study--Clubs 71.8 3130.9 11 6.53 18. Television 69.2 4644.7 8 8.65 19. Radio 33.8 1684.3 5 6.76 20. Newspapers 37.7 673.4 9 4.19 21. Magazines 47.2 872.4 11 4.29 22. Books 57.4 2276.9 9 6.38 23. Movies 160.0 6250.0 26 6.15 24. Social Activity 67.9 6186.1 7 9.70 25. Conversation 26.3 1057.4 4 6.56 26. Active Sports 96.5 4062.5 15 6.43 27. Outdoors 64.5 4250.0 8 8.06 28. Entertainment 132.6 4013.0 24 5.53 29. Cultural Events 77.9 1863.6 16 4.87 30. Resting--Naps 86.9 1835.3 19 4.57 31. Other Leisure 50.6 2376.8 7 7.23 206 is the opposite, skewed in the direction of the early morning hours, but from the opposite direction. Again the distribution is highly skewed, so the distribution is inverted, and subtracted from the last hour available in the activity selection matrix. It should also be mentioned again that while the activity duration statistics are described here in minutes, they are converted to hours with decimal fractions to facilitate computation. Activity Selection Statistics Estimating activity selection statistics and applying them is the simulation model presents one of the biggest problems for TIMMOD. To begin with activity selection data are simply not available in an ideal form. Data reported in The Use of Time are not aggregated for the total sample, but rather reported in three separate sub-classifications, employed men, employed women, and housewives. The activity categories are aggregated from the thirty-seven categories described before to eight. The eight categories include work, home and family (includes marketing and shopping), travel, eating, semi- leisure, television, other media, and other leisure. Unfortunately the aggregations for these tables do not coincide with the subtotals shown in the reduced thirty- seven activity code structure. In addition, the data are reported as the percentage doing a particular activity group, often with only one significant digit.34 207 It is simply not possible to estimate the selection statistics with the same precision as the activity duration statistics. The main reason for this, as discussed before, is the limitation imposed by the size of the sample in the United States Time Use Survey. A much larger study would be required to provide stable activity selection statistics. The estimation procedure for activity selection statistics begins with the combination of the three sub- classifications into a total sample estimate. This is accomplished by converting the percentages in each of the three sub-classifications to frequencies by multi- plying the total frequency by the appropriate percentage. The frequencies are then summed to provide total sample activity selection frequency estimates. Frequencies have been percented and are shown in Table 7. To expand the aggregate category estimates to the thirty-two TIMMOD activity categories a straight line estimation procedure is used based upon the overall time spent doing an activity. The activity selection matrix then becomes a twenty-hour by thirty-two activity categories matrix instead of a twenty-hour by eight activity category matrix. It is the expanded activity selection frequency matrix that is input to TIMMOD. Because of the loss of precision due to the lack of significant digits in the original selection matrices 208 m.m m.0 H.va H.0a 0.ma 0.0a 0.0 ~.m N.m H.0 0.0 0.5 0.m m.m 5.v 0.N 0.H 0.0 0.0 m.0 musmfimq Hmnuo 0.m «.5 0.0a m.HH m.ma 0.HH m.5 0.5 H.0 0.m H.m 0.0 0.m v.0 H.m 0.0 0.H 0.~ 0.0 H.0 mflvmz Hwnuo 0.5 0.va m.mm H.5m H.NN H.mH 0.5 m.m 5.v 0.v 5.v 0.0 m.~ H.m m.m 0.H 0.0 0.0 m.0 0.0 conH> toads H.v 0.0 v.~a 0.0a 5.0 5.0 ~.5 ~.m m.m 0.0 5.v 0.5 v.0 0.0 5.0 H.0 n.0H m.NH H.m v.H whamwmq . aflEmm 5.0 ~.~ 0.H m.~ ¢.m 0.~H 5.m~ H.HH m.~ 0.0 0.m v.HH H.0N 5.v m.m m.m 0.5 0.5 0.~ v.0 mcfiunm 0.H H.v m.m m.0 ~.0 0.5 ~.m m.va 0.0a H.m H.0 m.m 0.5 m.v 5.m m.v «.0 m.5 v.a ~.0 Ho>mu9 H.m 0.m 0.0 0.MH 0.0a 5.m~ m.m~ m.0m 0.0m 0.a~ 0.vm m.m~ m.v~ 5.5m 0.0m m.v~ 0.0m 5.mH 5.~ 5.0 >HHEmm a mEom m.m ~.m ~.0 m.0 0.0 ~.0 v.HH H.0H 5.mm 0.0a 0.Hv H.5m 0.v~ 0.Hv 0.~v v.Hv ~.0m «.ma 5.~ m.H xuoz Emma EQHH Emoa Ema 5mm 5&5 8mm Emm Emv Emm Emm Ema Emma Edaa Edoa Bum Sum EM5 Sow Edm wdo mmfi m0 maom wm mmHBH>HBU< 82mmmmmHD ZH Dmuduzm MUGBzmummm 5 mamde 209 and the rather imprecise matrix expansion procedure out— lined above, it is necessary to apply a differential weighting factor to the activity selection matrix. The weighting factor is simply multiplied by all of the fre- quency estimates for all hours of a particular activity category. Additional distortion in the activity selection matrix is caused by the logical connection of the four work-related activities. The weighting procedure helps correct for this. It should be noted that the weighting procedure applies only to the activity selection statis- tics and not to the activity duration statistics. The weighting is necessary because the required data are simply not available. Sample TIMMOD Output The TIMMOD output described here is the baseline run which will be used as the control condition in the simulation experiments which will follow in Chapter VI. The run shown is also the source of the TIMMOD output discussed in the model validation section of this chapter. TIMMOD output consists of a detailed printout including the number of occurrences, the average time spent, the variance for time spent, the percentage par— ticipating, and the time for those doing the activity for the thirty—two TIMMOD activity categories. This is shown in Table 8. An additional page of output is generated like the one for all participants for 210 TABLE 8 TIMMOD REPORT No. of Time . Percentage Time for Times Spent Variance Doing Those Doing 1. Main Job 2057.00 219.87 107499.49 49.50 444.18 2. Second Job 43.00 5.78 868.26 4.20 137.60 3. Work--Other 705.00 11.66 515.24 36.60 31.85 4. Travel to Job 1089.00 22.53 1684.63 43.40 51.91 5. Marketing 434.00 13.70 759.03 33.90 40.42 6. ShOpping--Errands 546.00 20.86 1422 87 41.80 49.90 7. Cooking 1660.00 43.56 3665 94 73.80 59.02 8. Home Chores 1988.00 58.79 6043 81 81.70 71.96 9. Laundry 653.00 24 28 1778 22 46.20 52.55 10. Pets and Garden 193.00 3.02 73.18 17.50 17.24 11. Other House 680.00 25.29 1812 58 48.10 52.58 12. Child Care 1405.00 36.35 2623 94 73.50 49.45 13. Personal Travel 2070.00 31.91 1873 58 82.70 38.59 14. Leisure Travel 1034.00 18.92 880.14 61.60 30.72 15. Eating 2992.00 84.13 10837 95 91.20 92.25 16. Personal Care 3039.00 70.49 7853.89 90.50 77.89 17. Study--C1ubs 414.00 28.85 3104 81 33.70 85.60 18. Television 1451.00 88.50 14360 05 75.90 116.60 19. Radio 108.00 3.14 112 64 10.30 30.46 20. Newspapers 672.00 24.63 1710 69 47.50 51.85 21. Magazines 167.00 7.40 419 24 15.70 47.11 22. Books 93.00 5.06 352 45 8.90 56.84 23. Movies 31.00 4.32 675.41 3.00 143.95 24. Social Activity 1064.00 67.28 9559.97 64.60 104.15 25. Conversation 668.00 16.52 807 35 47.80 34.57 26. Active Sports 65.00 5.83 624 47 6.20 94.04 27. Outdoors 39.00 2.28 149 78 3.90 58.46 28. Entertainment 50.00 5.77 802.09 5.00 115.45 29. Cultural Events 12.00 .90 73 67 1.20 74.68 30. Resting--Naps 239.00 19.88 2085.44 21.50 92.44 31. Other Leisure 428.00 20.30 1590.85 33.70 60.24 32. Night Sleep 1000.00 448.22 .00 100.00 448.22 211 participants in the single category specified on the run parameter card. A sample of the TIMMOD report for all participants is shown in Table 8. A TIMMOD summary report showing nine aggregate categories follows to simplify comparison with the experimental runs. Only the average time spent and the associated variance are shown in the summary. A sample of the TIMMOD summary is shown in Table 9. Model Validation As discussed in Chapter III, one of the most dif- ficult problems with computer simulation is in validating the computer model once it is develOped. Ideally, the multistage verification procedure recommended by Naylor et a1. should be employed with TIMMOD. At the heart of the multistage verification procedure outlined in Chapter III is the model's ability to predict the behavior of the system under study. Not only must TIMMOD replicate the sample statistics described by the United States Time Use Survey, but it is also required to demonstrate the ability to adequately predict changes in the allocation of time based upon changes in the structural composition of the environment, such as the addition of television or BCN technology. To completely verify TIMMOD, it must demonstrate the ability to handle changes through either historical verification or accurate forecasting. 212 TABLE 9 TIMMOD SUMMARY For All Individuals Time Spent Variance Work Related 259.83 110567.61 Shopping 34.56 2181.90 Home and Family 191.28 15997.67 Nonwork Travel 50.84 2753.73 Eating 84.13 10837.95 Semi-Leisure 99.34 10958.71 Television 88.50 14360.05 Other Mass Media 44.54 3270.43 Leisure 138.76 15693.62 213 Unfortunately the only validation possible with TIMMOD is in the model's ability to replicate the original data. It is not possible to verify the model through any historical exercise. TIMMOD might be verified by comparison with the 1935 time-budgets of Sorokin and Berger. It would seem simple to remove television from TIMMOD, which it is, but television is not the only dif— ference between 1935 and 1965. For example, there have been tremendous changes in the transportation system indicated by improved highways and automobiles, not to mention the population growth, the move to the suburbs and shopping centers. The differences between 1935 and 1965 are simply too complex to describe in terms of changes in a few global activity categories. The same is true for a cross-national comparison. The only validation that is possible for TIMMOD is the demonstration of the model's ability to reproduce the original data. Remember that TIMMOD has reduced the original data to a series of theoretical probability distributions and stochastic processes. If the model can replicate the original data, then there is some basis for an argument that the mathematical manipulations employed by the model are indeed valid. Three important kinds of data generated by TIMMOD are considered here. First is the total number of occurrences of a particular activity for the number of 214 cycles run using the model. In this case there are 1,000 cycles or simulated individuals. The original data number of occurrences per 1,000 individuals is calculated by multiplying the percentage participating in a particular activity by the number of occurrences per participant by 1,000. This comparison is shown in the first two columns of numbers in Table 10. A Pearson product-moment correlation is then calculated as an index of how well the simulation output matches the original data. In this case, r = .998. This high correlation should be expected, however, because of the weighting scheme employed on the activity selection statistics. The most critical comparison is in the average time spent for the total sample. This measure is critical because it is the criterion variable to be used in the simulation experiments in Chapter VI. For the thirty-one mean times spent shown in the third and fourth columns of Table 10, the r = .999. Another comparison is shown between the percentage participating in the original data, and the TIMMOD output. It should be mentioned that there is no logical con- nection between the various activity categories, except the work-related activities. For the percentages repre— sented in the fifth and sixth columns of Table 10, the 2115 TABLE 10 COMPARISON OF THE ORIGINAL DATA AND TIMMOD OUTPUT Occurrences per Overall Average Percentage Activity 1000 Individuals Time Spent Participation Study TIMMOD Study TIMMOD Study TIMMOD 1. Main Job 2048 2057 225.0 219.9 51.2 49.5 2. Second Job 36 43 4.9 5.8 2.4 4.2 3. Work--Other 699 705 11.5 11.7 33.3 36.6 4. Travel to Job 1217 1089 24.9 22.5 50.7 43.4 5. Marketing. 447 434 14.0 13.7 34.4 33.9 6. Shopping-Errands 457 546 17.9 20.9 30.5 41.8 7. Cooking 1636 1660 43.8 43.6 60.6 73.8 8. Home Chores 1949 1988 58.4 58.8 60.9 81.7 9. Laundry 678 653 25.5 24.3 29.5 46.2 10. Pets and Garden 195 193 3.3 3.0 12.2 17.5 11. Other House 641 680 23.7 25.3 37.7 48.1 12. Child Care 1278 1405 32.6 36.4 37.6 73.5 13. Personal Travel 1969 2070 30.8 31.9 63.5 82.7 14. Leisure Travel 1073 1034 19.3 18.9 42.9 61.6 15. Eating 2857 2992 81.3 84.1 98.5 91.2 16. Personal Care 2946 3039 ‘ 68.5 70.5 98.2 90.5 17. Study--Clubs 386 414 27.7 28.9 20.3 33.7 18. Television 1324 1451 91.6 88.5 69.7 75.9 19. Radio 107 108 3.6 3.1 8.2 10.3 20. Newspapers 632 672 23.8 24.6 48.6 47.5 21. Magazines 136 167 6.4 7.4 11.3 15.7 22. Books 92 93 5.3 5.1 7.1 8.9 23. Movies 20 31 3.2 4.3 2.0 3.0 24. Social Activity 928 1064 63.0 67.3 54.6 64.6 25. Conversation 700 668 18.4 16.5 41.2 47.8 26. Active Sports 57 65 5.5 5.8 5.7 6.2 27. Outdoors 37 39 2.4 2.3 3.1 3.9 28. Entertainment 41 50 5.4 5.8 3.7 5.0 29. Cultural Events 8 12 0.6 0.9 0.7 1.2 30. Resting--Naps 223 239 19.4 19.9 18.6 21.5 31. Other Leisure 386 428 19.5 20.3 25.7 33.7 216 r = .946. Needless to say all of these correlations (with n=31) are statistically significant. While TIMMOD demonstrates validity with no experimental manipulation, any generalization of TIMMOD to experimental conditions should be done only with extreme caution. TIMMOD is limited by design to be a forecasting tool. It is intended only to estimate changes in the allocation of time based upon the manipu- lation of a few hypothetical environmental conditions, such as capability of BCN technology. It is not intended to become a general model for the human allocation of time. It is rather a forecasting tool allowing vicarious experimentation which is meant only to improve on the armchair scenario. CHAPTER V- -NOTES 1Richard K. Brail, "Activity System Investi- gations: Strategy for Model Design" (Ph.D. dissertation, University of North Carolina, 1969), pp. 136-46. 2George C. Hemmens, "Analysis and Simulation of Urban Activity Patterns," Socio-Economic Planning Sciences 4 (1970): 56. 3 Brail, "Activity System Investigations,‘ pp. 142- 44. 41bido I p. 1450 5User's Guide: The 6500 Scope/Hustlergperating System, Vol. II, Computer Laboratory, Michigan State University, East Lansing, Michigan, pp. 10.10-10.27. 6FORTRAN-Extended Reference Manual, Control Data Corporation, Sunnyvale, California, 1973. 7A. Alan B. Pritsker and Philip J. Kiviat, Simulation with GASP II (Englewood Cliffs: Prentice- Hall, 1969), p. 15. 8Thomas H. Naylor, Joseph L. Balintfy, Donald S. Burdick, and Kong Chu, Computer Simulation Techniques (New York: John Wiley & Sons, 19665, p. 69. 9Ann Hughes and Dennis Grawoig, Statistics: A Foundation for Analysis (Reading: Addison-Wesley, 1971), p. 2. loNaylor, Balintfy, Burdick, and Chu, Computer Simulation Techniques, p. 69. 217 218 11Chris P. Tsokos, Probability Distributions: An Introduction to Probability Theory with Applications (Belmont, Calif.: Wadsworth Publishing Company, 1972), p. 95. lzIbid., p. 142. 13Peter W. Zehna, Probability Distributions and Statistics (Boston: Allyn and Bacon, 1970), pp. 122-60. l4Ibid., p. 133. 15Tsokos, Probability Distributions, p. 113. l6Zehna, Probability Distributions and Statistics, p. 141. 17Naylor, Balintfy, Burdick, and Chu, Computer Simulation Techniques, p. 81. 18Zehna, Probability Distributions and Statistics, p. 148. 19Tsokos, Probability Distributions, pp. 151-83. 20E. Martin Basic, "Development and Application of a Gamma-Based Inventory Management Theory" (Ph.D. dissertation, Michigan State University, 1965), P. 8. 2:I'Nay-lor, Balintfy, Burdick, and Chu, Computer Simulation Techniques, p. 45. 221bid., p. 46. 23FORTRAN Extended Reference Manual, p. D-4. 24Pritsker and Kiviat, Simulation with GASP II, pp. 96-102. 25 Naylor, Balintfy, Burdick, and Chu, Computer Simulation Techniques, pp. 70-73. 26Ibid., pp. 87-88. 219 27Ibid., p. 88. 28Summary of United States Time Use Survey, Institute for Social Research, University of Michigan, Ann Arbor, Michigan, 1966. 29Alexander Szalai, ed., The Use of Time (The Hague: Mouton, 1972), pp. 495-825. 30Ibid., p. 580. 311bid., p. 581. 321bid., p. 582. 33Ibid., p. 822. 34Ibid., pp. 746-71. CHAPTER VI THE SIMULATION EXPERIMENT Experimental Paradigm Experimentation with TIMMOD involves the manipu- lation of various probability distribution input para- meters as independent variables with the concomitant variation in the allocation of time to various activity categories as the dependent variable. TIMMOD has been designed to provide a means of performing the vicarious experimentation discussed in Chapter III which may be appropriate to the assessment of the future impact of structural changes in the environment, such as new com- munication technology. The experiment outlined here is intended to provide insight into the future impact of BCN technology in a forecasting sense. The experiment is not addressed to specific hypotheses designed to test any marketing communication theory, but rather is sug- gested as a potential methodology for technological forecasting. The experimental design employed in TIMMOD experiments is a simple before and after design comparing 220 221 the allocation of time to various activity categories before and after experimental manipulation. A control condition is not necessary for reasons which will be dis- cussed later. To simplify analysis the dependent variable is restricted to nine major categories of daily human activity as described in Chapter V. The major cate- gories represented are work-related, shopping, home and family, nonwork travel, eating, semi-leisure, television, other mass media, and leisure. These nine categories represent an aggregation of the thirty-two categories described as the basic TIMMOD output to simplify the process of comparison. Work-related activities are an aggregation of main job, second job, at work other, and travel to job. Shopping consists of marketing, and shopping and errands. Home and family consists of cook- ing, home chores, laundry, pets and garden, other house and child care. Nonwork travel consists of personal travel, which includes travel for both marketing and shopping, and leisure travel. Eating category is the same. Semi-leisure includes personal care, and edu- cational and organizational activity. Television cate- gory is the same. Other mass media includes radio, newspapers, magazines, books, and movies. Leisure activities include social activity, conversation as a primary activity, active sports, outdoor activities, entertainment and amusements, cultural events, resting 222 and naps, and other leisure. Marketing is used here to describe shopping behavior for everyday nondurable goods, and is the term adopted by the United States Time Use Survey. Shopping in the original thirty-two categories means only shopping for durable consumer goods. In the aggregated shOpping category, shopping means shopping for all consumer goods. Later in this chapter the distinction is made between the term "shopping,' which refers to shopping for only durable consumer goods, and all shopping which combines both shopping for durable goods and market- ing, or shOpping for nondurable goods. The dependent variablethen becomes the allocation of minutes for the average employed adult to the nine activity categories. The independent variables center around the theoretical probability distribution described in Chapter V. Three parameters can be manipulated, the a parameter of the Erlang distribution which describes the number of minutes per interval or the skewness of the distribution, the B parameter which is the reciprocal of the average number of minutes per interval, and the weighting factor necessary to adjust the activity selection statistics discussed in Chapter V. Manipu- lating the a parameter would have the effect of making the activity duration distribution more or less skewed; manipulating the B parameter would have the effect of making the activity duration distribution being longer 223 or shorter per unit interval, that is a longer or shorter average duration, and manipulating the weighting factor would have the effect of making the activity being selected either more or less often throughout the entire day. The a parameter is not manipulated here, that is the shape of the activity duration distributions, because there is not sufficient evidence to build any case for this type of impact. The shapes of the activity distri- butions are then assumed to be the same. The weighting factor is also not manipulated, nor are any of the basic activity selection statistics, because the empirical base for these statistics is sufficiently weak to generate some lack of confidence in their ability to represent activity selection throughout the day. Because of this lack of confidence in the activity selection statistics, they are held constant. The only parameter that has been manipulated as an independent variable is the B parameter which has the effect of lengthening and shortening the average duration of a particular activity event. All experimental manipu- lations using TIMMOD that are described here involve changes in the B parameter. It should be pointed out that the dependent variable is the result of the inter- action of both the activity selection process and the activity duration process. The activity duration 224 parameters describe only the duration of a single activity event, and the dependent variable describes the total time devoted to the particular activity category for the entire day. The specific treatment conditions and their translation to the B parameter are described later in this chapter. Experimental and Statistical Problems in Using TIMMOD Experimentation with computer simulation models, such as TIMMOD, involves making alterations either one at a time or in combinations in the model and noting the effects of these alterations. ‘While at first glance this may appear relatively simple, statistical techniques which are traditionally applied in such experimental situations are no longer sufficient in the computer simu- lation situation. As Jacoby and Harrison point out, because computer simulation models possess Special char- acteristics, it is often necessary to extend and modify the classical techniques.1 One of the first problems to be considered with most simulation models is the size of the model as dis- cussed in Chapter III. Often computer simulation models grow to a very large size, with large numbers of both input and output variables making it extremely difficult to design manageable experiments using the model. How- ever, this is not a critical problem with TIMMOD because 225 the model is relatively small and simple by comparison to other simulation models. Essentially there are only three possible input variables and only one output variable. The output variable has been somewhat simplified by aggregating the original thirty-two cate- gories to the nine categories described earlier in this chapter. While other output is possible, such as fre- quency of occurrence or participation rate, this analysis will concentrate on the total time allocated to various activities. While TIMMOD avoids most of the experimental problems caused by size, it can be argued that it looses isomorphism with the real world because of oversimplifi- cation. This problem will be discussed again later. Jacoby and Harrison describe two special features of simulation models. The first is that a simulation model can be considered a closed system, and the second is that the random number sequences are strictly deter- mined.2 A closed system should contain no surprises or unknowns that might be expected in an Open real-world system. In some circumstances this might be highly desirable experimentally, but in technological forecast- ing it is somewhat disturbing. Technology is commonly not viewed as a closed system because of the unexpected discovery such as the laser as described in Chapter III. However, a closed system limits the range of potential outcomes the system can have. The implications for 226 TIMMOD experiments involve assumptions that future activities will be selected in approximately the same manner as the present and that new factors which could play a role in the selection and duration of human activ- ities would not be introduced. The fact that random number sequences are strictly determined in computer simulation through the use of psuedo-random number generators as discussed in Chapter V makes the control of error variance possible. Error is known precisely. It is possible to manipulate error variance through the use of what Jacoby and Har- rison term the Monte Carlo swindle.3 If each experimental run is started with the same random number seed, then the sequences will be the same, generating an artificial correlation between the runs, which would have the effect of suppressing error variance. This obviously must be taken into account in designing simulation experiments. The TIMMOD experiments all begin with the same random number seed, thus taking advantage of the swindle. Fishman describes a related problem in gener- ating time series from computer simulation experiments in terms of autocorrelation.4 Because of the way random numbers are generated in computer simulation experiments, special consideration must be given to assumptions about statistical independence in experimental designs. Another problem involves the stability of the activity selection and duration statistics over varying 227 treatment conditions. This problem was discussed in Chapter IV. It is possible, as Tomlinson, Bullock, Dickens, Steadman, and Taylor point out, that time allo- cations may well change under different physical con- ditions within the environment, enough, to seriously limit the generalizability of a model based on a single set of time allocation data.5 This caveat should be kept in mind when interpreting TIMMOD results. Analysis Method The basic comparison made in the TIMMOD experi- mental runs is the difference in time allocated to various activity categories in minutes. The comparison can be for any single activity category, such as con- sumption of mass media, or for all activity categories considered as a whole. The consideration of impact of BCN technology is in terms of apparent differences in this time allocation process. For example, if mass media had fewer minutes allocated to it for the entire day, then the implication would be that this is evidence of the impact the particular experimental treatment. In this case, experimental treatments are assumed to reasonable estimates of the result of the introduction of a BCN system. The selection of a statistical technique to evaluate the differences in time allocation must begin with careful consideration of the time allocation data. 228 First is the problem already discussed of the before and after conditions not being statistically independent. This requires the use of a correlated measures design to reap the full benefits of the lack of independence. The use of an independent measures design would not take full advantage of the correlation, and thus would tend to overstate the error, making the procedure substan- tially less sensitive to any change in time allocation. A second problem involves the manner in which time is allocated to the various activity categories. The waking and retiring times are determined indepen- dently of other activities which creates a standard length day to fill with various activities. Since the length of the day remains constant within sampling error tolerances, the sum of the days' activities always remains approximately the same regardless of any change in the distribution of activities. In other words, the activity categories are not independent. This presents a problem in applying traditional analysis of variance techniques because the difference across runs would always be the same, that is within the sampling error of the length of the day. Deve10pment of statistical techniques to accommodate these problems is beyond the scope of this dissertation. Also, the intent is to demonstrate a methodology applicable to technological forecasting 229 rather than to develOp theory. The statistical analysis used in reference to differences in time allocations will be both simple and straightforward. The reason for employing any statistical measures at all is to provide a means of comparing the magnitude of differences across the various treatment conditions. The difference between the average time allocated to a particular activity across the TIMMOD runs is shown as both the raw difference in minutes and as a standard score. The difference measure is standardized by using a simple comparison of different means equation as follows: Xb - Xe z _ 2 2 /be/1000 + Se/1000 The mean time allocated to the particular activity with- out the experimental treatment is shown as Xb and the mean time allocated with the experimental treatment is Xe. The variances for the two groups are S: and s: respectively. Both groups are for 1,000 simulated indi- viduals. This standard score then provides an index to evaluate the magnitude of the difference in time allo- cation in a given single category. The standard score estimate is conservative because it does not take into account any correlation between the two groups. The raw data are not readily available to calculate the necessary 230 correlations from the TIMMOD runs. The use of the standard score here is intended only as an estimate of the magnitude of the difference between the two means as compared to what might be expected if chance were allowed to Operate. It is not sufficient to describe the difference in terms of minutes alone. For example, a five-minute difference in shopping would indicate a greater degree of impact than would a five-minute dif- ference in work-related activity. The reason is the relative magnitudes of the two activity categories, shopping requiring approximately 35 minutes per day and work—related activity accOunting for nearly 260 minutes per day. The use of the standard scores here is limited to an index of difference between single pairs of activities. No attempt is made to ascribe statistical significance to any pair of activity cate- gories. While the standard scores can be used to describe the relative magnitude of difference between any set of single pairs of activity categories, it is not possible to describe the overall change for all the activity cate- gories in this manner. Remembering that the sum of all of the activities for each experimental run is approxi- mately the same, the comparison of interest is in the difference in shape of the two distributions. This suggests that minutes allocated to a particular activity 231 category can be represented as a frequency and that a chi-square goodness-of-fit test can be applied. The use of the chi-square is suggested by Tomlinson et a1.6 McNemar describes the goodness-of—fit test as follows: If we wish to check on whether it is reasonable to believe that a given frequency distribution is, within the limits of chance sampling, of the normal or some other specified type, a frequency curve having the same basic constants as those computed from the observed frequency distribution can be fitted to the data.7 In this case, the no-treatment TIMMOD run becomes the theoretical probability distribution, and the experi- mental run becomes the observed frequency distribution being fitted. The goodness-of-fit statistic is calcu- lated as follows: 2 2 (Mb - Me) x = Z M e The chi-square statistic is calculated from the number of minutes in each nontreatment activity category, Mb, and the number of minutes in each treatment activity category, Me. The goodness-of-fit chi-square is used in much the same way as the standard scores for each individual pair. This statistic serves as an index for the overall impact of the treatment condition across all of the activity categories. Again, statistical significance is not important. 232 Each TIMMOD run is based upon a simulated sample of 1,000 individuals. This sample size was selected primarily because of computational convenience. A larger sample size or number of cycles was not selected because it would contribute to misleading precision. The original United States Time Use Survey consisted of only 1,244 individuals. The sample selected was not smaller to minimize the problem of stochastic convergence discussed in Chapter III. Experimental Conditions The experimental conditions used to represent the features of a BCN system include transportation related to shopping, television viewing, and the time spent shopping. Each of the conditions are represented in terms of an increase or reduction of the B parameter of the appropriate Erlang distribution as previously discussed. The increases and decreases are represented as a low and a high value to simplify the analysis. Each low value is represented by a 10 percent increase or decrease in the B parameter while each high value is represented by a 50 percent increase or decrease. Transportation Transportation is the amount of time saved in making shopping trips as a result of having a BCN system. The reduction in the 8 parameter is in the personal 233 transportation category, which includes other transpor- tation such as travel to accompany children, travel necessary to purchase goods and services, and travel necessary for personal needs. Shopping and marketing related travel accounts for about 60 percent of all of the travel time in the category.8 The B parameter for this distribution is 5.2 minutes and the a parameter is 3. In the nontreatment condition this provides a mean of 15.6 minutes and a variance of 81.1 minutes. In the low transportation treatment, representing a 10 percent reduction in the B parameter,the B parameter is decreased to 4.7 minutes providing a mean of 14.0 minutes and a variance of 65.7 minutes. In the high transportation treatment, the B parameter is decreased 50 percent pro- viding a B parameter of 2.6 minutes and a mean personal travel time of 7.8 minutes and a variance of 20.2 minutes. Additional discussion of transportation as a key variable is assessing the impact of BCN technology is found in Chapter II. Television Television is the amount of time spent viewing television as a result of having the BCN system which would make viewing television more attractive because of the addition of interactive programming and a broader program offering. In this treatment, the B parameter is increased to represent an increase in viewing time to 234 accommodate the more attractive program offering. In the low television treatment, there is a 10 percent increase in the B parameter. This increase is also discussed in Chapter II. In the nontreatment television Erlang distribution the B parameter is 8.7 minutes and the a parameter is 8. This provides a mean viewing time per television viewing event of 69.2 minutes and a variance of 598.6 minutes. In the low television treat- ment the B parameter is increased to 9.5 minutes pro- viding a mean of 76.1 minutes and a variance of 723.5 minutes. In the high television treatment, which is a 50 percent increase in the B parameter, the B parameter is increased to 13.0 minutes, providing a mean of 103.8 minutes and a variance of 1,347.8 minutes. Marketing The marketing category includes the time allo- cated to shopping for nondurable or everyday goods on a per event basis. The B parameter for this Erlang distribution is 6.3 minutes and the a parameter is 5. The nontreatment B parameter for marketing is 31.3 minutes with a variance of 195.8 minutes. Following Bucklin's reasoning as described in Chapter II, as shopping becomes less expensive, the time allocated to it should increase since consumers have a tendency to shop more. The low marketing treatment, then, is represented by a 10 percent increase in the B 235 parameter. The low marketing treatment B parameter is then 6.9 minutes providing a mean of 34.5 minutes and a variance of 237.2 minutes. The high marketing treat- ment is represented by a 50 percent increase in the B parameter giving a B parameter of 9.4 minutes. This provides a mean of 47.0 minutes and a variance of 440.9 minutes. Shopping The shopping category includes the time allocated to shopping for durable consumer goods, and is manipulated in much the same way as is the marketing category. The nontreatment shopping B parameter is 7.8 minutes and the a parameter is 5. This provides a nontreatment mean of 39.1 minutes and a variance of 305.6 minutes. The low shopping treatment, again represented by a 10 percent increase in the B parameter, yields a B parameter of 8.6 minutes. This provides a low shopping treatment mean of 43.0 minutes and a variance of 369.6 minutes. The high shopping treatment is a 50 percent increase in the 8 parameter, or a B parameter of 11.7 minutes. The high shopping treatment yields a pershopping event mean of 58.7 minutes and a variance of 688.0 minutes. The all shopping treatment is represented by the combination of both the marketing and shopping treatments. Combinations of the various treatments are shown later in this chapter. 236 Limitations There are a number of limitations which should be considered when interpreting the results of TIMMOD experiments. All the limitations are related either to simplifying assumptions made in the early stages of the model development, to the lack of a broad empirical data base, or to the very difficult problem of model verifi- cation. TIMMOD assumes that the cable penetration will be 100 percent of all households, or at least all house- holds represented in the United States Time Use Survey. No attempt was made to assess the varying impact due to different levels of penetration, which could be done by multiplying the penetration percentage by the results shown here. Also, it is assumed that a BCN subscriber would be demographically the same as the average American, which is probably not the case. As discussed earlier, an adequate data base is not available to develop all of the probability distri- butions for all activities for each hour of the day. It is also not possible to develop the probability distributions for very many different population seg- ments, such as employed men, employed women, housewives and children. A study which would provide adequate data would be extremely large and very expensive. The weakest data input into the model here is the activity 237 selection statistics. This problem is discussed in more detail in Chapter V. Also, the activity duration sta- tistics remain constant throughout the day. Related to the data problem is the fact that there is no logical connection between various activity categories, except working. The reason for this is insufficient data to describe the logical connections. A priori "commonsense" logical connections were considered, but could easily create more distortion than no logical connections at all. The only relationship between activity categories is the time-of—day probability of occurrence. It is felt that odd combinations of activi- ties which may result in a single individual's activity allocations will average out across a large number of individuals or cycles. No attempt is made to make any statement about a single individual's time allocations. Better specification of treatment conditions is also desirable, but again there is simply not sufficient evidence to justify it. More than just the B parameter can be experimentally manipulated using TIMMOD. However, studies do not exist which describe shapes of distri- butions and time-of—day allocations of various activities under the conditions which can be anticipated with the application of BCN technology. No doubt the various demonstration projects using BCN technology will provide some of the necessary estimates. 238 Perhaps the most serious limitation of the TIMMOD experiments is the difficulty encountered in verifying the TIMMOD model itself. Since TIMMOD is intended as a technological forecasting tool, its ultimate verification is in its ability to forecast the future impact of BCN technology. This problem is discussed in both Chapter III and Chapter V, and a satisfactory solution to the problem is simply not possible. Another problem, described by Fishman, is the lack of analytical models develOped to handle the special cir- cumstances involved in computer simulation experiments.9 The statistical techniques necessary for the analysis of this kind of data will begin to appear in the literature. Impact of Transportation The results of the low and high transportation treatment experimental runs are shown in Table 11 and Table 12. In the low transportation condition, the overall reduction in nonwork travel is only 2.8 minutes. This means that while each trip is shorter, more trips are made. The average adult still spends 48.0 minutes per day engaging in nonwork travel. The overall impact of the low reduction in personal travel is virtually nil. The chi—square for the low transportation experiment is .87, which translates to a nonchance probability of less than .005 at eight degrees of freedom. 239 mamm.- m.H- m.5~mma c.5mH m.mmmma m.mMH musmauq mmmm. m.~ s.m5sm p.04 v.05mm m.s¢ mapmz mum: H045.- m.m- m.msama p.4m o.oomsa m.mm acnmn>mame 0000. o. m.mmmoa m.mm S.mmmoa m.mm musmnmq-HEmm 0000. o. 4.5msoa H.4m o.mmmoa H.4m manumm mmmm.a- m.~- m.~4¢~ o.ma a.mm5~ m.om Hu>mue xuozcoz sppm.- m.H- 6.6H5ma m.ama n.5mmma m.HmH Saaamm can meow mm5m.u 8.H- m.m5om m.mm m.HmHm 6.4m mcflddoam 5mam. m.m m.omm5HH H.mmm p.5mmOHH m.mm~ pmumamm xuoz mUGMAHm> umwmm mocmwnm> “MMMM muommumo muoomIN mononmwwfio muw>fluom cum Hmpcmfiflummxm cum Honucou somdzH oneaemOdmzame son quzomm maqsmmm oozzHe Ha mqmdB 240 m5mm. m.a a.0000H 5.0va 0.000ma 0.0ma ousmwmq H05v.n N.H| 0.000m m.mv v.05mm m.v¢ mwpmz mmmz mmmm.n 0.0: 0.0vaH m.m0 0.0mmva m.00 sowmfi>mame 550H.u m.u m.5mm0a 0.0m 5.0mmoa m.mm madmflmquwemm HN5H. 0. m.H050H 0.00 0.0000H H.v0 .mcfiumm 50m5.5| 5.mHn 0.vaa H.mm 5.mm5~ 0.0m Hm>mne xuozcoz 00HH.H v.0 5.0005H 5.50H 5.500ma m.HOH maflsmm paw mfiom mmmm.au 5.N- 5.mmma m.Hm m.HmH~ 6.4m maaddoam mmmm. 0.m 0.mmmmHH 0.m0m 0.50m0HH 0.0mm pmumamm xnoz mocmwum> uwmmm mucmwnm> uwwwm muomwumu whoomum mocmummmwo huw>fiuo¢ cum Hmucmswummxm cum Houucoo BufimZH ZOHBdBmOmmzuaue sm5~.- m.H- p.5mpoa o.mm 5.mmmoa m.mm musmfiuqun88m mamm.- m.m- o.op5m ~.om o.mmmoa H.8m mcaumm Hmmm. m. o.mvmm G.Hm 5.mmhm m.om Hm>mua xuozcoz mmpm.- N.m- 5.mmpmH H.mmH 5.5mmmH m.HmH Sausam 6cm meow Hmmm.u 5.H- 5.Hmma m.~m m.HmH~ 0.4m maflddosm memo. m. m.5mmoaa H.omm p.5pmoaa m.mmm umumamm x803 mocmwum> DMWMM wocmwum> DMHMM whommumo muoomlw mocmanMHo mufl>auom com Hmucmswnmmxm cam Houuqoo BoamzH onmH>mqme son quzomm meqsmmm dozzHe ma mqmde 243 504m.au 5.0a- ~.5qufl H.m~H p.mmmmH m.mma musmnuq mmmm.au m.s- m.mm5~ u.mm 4.05mm m.ss mapwz mum: omm~.s H.6N «.4mflmm 0.4HH o.ommvfl m.mm conn>mHue ammo.al 5.4- m.oamm 6.4m 5.mmm0H m.mm musmnmqunsum mmom.- 4.4- m.5mmm 5.m5 o.mmm0H H.4m mcaumm mmaa.- H.~- m.m~m~ 5.ms a.mm5~ m.om Hm>mue xuozcoz m5mm.H- m.m- 5.mHHmH o.~mH 5.5mmma m.HmH Saaemm ppm meow omm5.u G.H- ~.5¢Hm o.mm m.HmHm 0.4m ucflddoam mmma.- m.~- G.O5ma0H 0.5mm p.5omoaa m.am~ pmumfimm xuoz mocmwum> DMMMM mocmflum> uwwwm whommumo onoomuN wocmummmwo muw>wu04 cam Hmucmfiwummxm cam Houucoo eoddzH onmH>mnma mon oszomm maqammm nozzHe «H mgmdB 244 The high television condition is a different situation. The high television treatment provides a 26.1 minute increase in television viewing time for a total of 114.6 minutes per day. The chi-square for the high television experiment is 10.12 which translates to a nonchance probability of nearly .75. The greatest impacts are in the reduction of 10.7 minutes from the leisure activity category, a reduction of 4.9 minutes in the other mass media category, and a reduction of 9.3 minutes in the home and family category. Also, there are small reductions in nonwork travel, eating and semi-leisure. The high television condition appears to have greater impact than does the high transportation condition. Television alone, then, appears to have the greatest impact on other mass media and leisure in general, with a somewhat smaller impact on home and family activity. Television time is a very important dimension of the overall impact of BCN technology. Impact of Shopping Because of the division of shopping into two cate— gories, marketing or shopping for nondurable goods, and shopping for durable goods, the consideration of the impact of shopping is somewhat more complex. Marketing, shopping, and all shOpping will be considered in turn. The low marketing condition results in an overall decrease in marketing activity of .8 of a minute. 245 This is because a small increase in marketing event time apparently causes a reduction in the number of shopping events occurring during the day. The chi-square for the low marketing condition shown in Table 15 is 1.01, which translates to a nonchance probability of less than .005. While the low marketing condition seems to show a reduction of 6.4 minutes in eating, overall it has vir- tually no impact. The high marketing condition is very similar. As shown in Table 16, the high marketing con- dition increases the overall shopping time per day 5.1 minutes to a total of 39.7 minutes. The chi-square for the high marketing condition is 1.03, translating again to a nonchance probability of less than .005. An increase in the attractiveness of shopping for non- durable goods, then, appears to have little impact on the allocation of time to other activities during the day. Low and high durable good shopping are shown in Table 17 and Table 18. The low shopping condition results in a 1.0 minutes decrease in overall shOpping to 34.6 minutes per day. The chi-square for the low shopping condition is 1.20, which is also translated to a nonchance probability of less than .005. At least in the low shopping condition, like marketing or non- durable shOpping, there is virtually no impact whatsoever. In the high shopping condition, however, a very slight trend begins to emerge. The high shopping 246 mpoom Hmfismcoo mHQMHSpco: mom mcwmmonm k. ommo. N. c.5pmmH o.ama o.mmmma m.mma musmnmq momm. o.H H.5mmm m.mv S.O5~m m.sv mead: mum: Ho~5.- m.m- «.5msma 5.4m o.oomva m.mm coamfi>uame mmma.- 5.- 0.55moa q.mm 5.5mmOH m.mm musmamq-flsmm O5~v.H- a.o- p.m5~m 5.55 o.mmmoa H.4m mcflumm «6mm. 0. 4.0555 ¢.Hm 5.mm5~ m.om Hm>mue xuozcoz 5005.- m.H- m.msmma m.mmH 5.5mmmH m.HmH 5Hflemm 0cm 880m mmmm.- 5.- 4.H~H~ m.mm m.HmH~ p.8m mcaddoam 555m. 5.m 5.45omHH m.mm~ p.5mmoafi 5.5mm pmumfium x803 mocmflnm> uwwwm mOGMHHm> DMWMM muommumo muoomIN mocmummmao mufi>wyom cam HmucmEauwmxm cam Houucou soddzH «GZHBmMMm0me 0000. 0.0 0.00000 0.000 5.00000 0.00 musm0mql05mm 0055.- 0.0- 5.00500 0.50 0.00000 0.00 000000 0000. 0. 0.5050. 5.00 5.0050 0.00 0m>mua xno3coz 0000.: 0.0I 0.00000 0.000 5.50000 0.000 >00Emm can 080m 0005.5 0.0 5.0055 5.00 0.0005 0.00 00000000 0000.1 5.01 0.000000 0.000 0.500000 0.000 pmu00mm xnoz mUCM0Hm> umwmm mUCM0Hm> uwwmm anommumu muoomum moamummm0o mu0>0uo¢ com 0Mpcmfi0nmmxm cam 0ouuaoo aommzH meqammm 002209 «UZHBHMfidS mUHm UZHzomm 00 mflmfia 248 00000 umESmcoo 00nmnsw Mom 000mmosm .1 5500. 0. 5.00000 0.000 0.00000 0.000 0000000 5000. 0.0 0.5050 0.00 0.0550 0.00 00002 0002 0500.- 0.5- 0.05000 5.00 0.00000 0.00 000m0>0000 0000.- 5.0- 5.00500 0.00 5.00000 0.00 0000000-0500 0005.- 0.0- 0.05000 0.00 0.00000 0.00 000000 5005. 0. 5.0055 . 0.00 5.0055 0.00 00>0ua 0003002 0000.- 5.5- - 0.00000 0.000 5.50000 0.000 000200 000 0200 5550.- 0.0- 0.0055 0.00 0.0005 0.00 00000000 0000. 5.00 0.055000 0.055 0.500000 0.005 0000000 0003 OOGMfiHm> #MWMM OOCMflHMNV “WWW...“ \aHOUOMMU 00000-0 mocmummm0o >u0>0uo¢ cam 0mucmE0mexm cam 0ouucoo BUdAZH «02Hmmomm 300 UZszmm mBQDmmm QOZZHB 50 mnmda 249 00000 Hmsdmcoo 0030056 00m mc0mmonm t. 0000.- 0.0- 0.50000 0.000 0.00000 0.000 0000000 0055. 5. 0.0550 5.00 0.0550 0.00 00002 0002 0500. 0.0 0.00500 0.00 0.00000 0.00 00000>0000 0000.- 0.0- 5.00000 5.00 5.00000 0.00 0000000-0500 0555. 0.0 5.00000 0.00 0.00000 0.00 000000 0050.- 0.- 5.5005 0.00 5.0055 0.00 00>000 0003002 0000. 5. 0.00000 0.000 5.50000 0.000 000000 000 0200 0000.0 0.0 0.0050 0.00 0.0005 0.00 00000000 0055.- 0.00- 0.000000 5.005 0.500000 0.005 0000000 0003 mommanm> UMMMM mOGMHHm> uwmwm %Homwumo whoomlN mocmummmwa >uw>0uo< cam 000:080ummxm cam 0ouucou BodmZH mH mqmde «OZHAAONw EUHE DZHSOmm mBADmmm DOSEHB 250 condition demonstrates a 9.0 minute increase in shopping to 43.6 minutes per day. The chi-square for this experi- ment is 3.10 which translates to a nonchance probability of less than .10. There is a very slight indication that semi-leisure activity is the most vulnerable to this increase, with an overall reduction of 4.6 minutes. The balance of the time is spread across the other activity categories. Combining the nondurable and durable shopping categories provides more interesting results. Shown in Table 19 are the combined results of low marketing and low shopping conditions. This results in a 1.1 minute reduction in shopping time, but interestingly a dis- turbance in the overall allocation of time showing a 5.0 minute reduction again in semi-leisure time. The chi-square for this experiment is 1.25, and again trans- lates to a nonchance probability of less than .005. So far there has not been any demonstrable impact of the overall allocation of time as a result of any change in shopping behavior. The combined high marketing and high shopping begin to show some impact as shown in Table 20. The high all shopping condition provides a 13.6 minute increase in shopping for a daily total of 48.2 minutes. The chi-square here is 9.79 which is converted to a nonchance probability of less than .75. The trend of 251 5000.- 0.- 0.00000 0.000 0.00000 0.000 0000000 0000.- 0.0- 0.5000 0.00 0.0550 0.00 00002 0002 0000.- 5.0- 0.00050 0.00 0.00000 0.00 00000>0000 0000.0- 0.0- 5.00000 0.00 5.00000 0.00 0000000-0000 0500.- 5.5- 0.00500 0.00 0.00000 0.00 000000 0505. 0. 5.0505 0.00 5.0055 0.00 00>000 0003002 5055.- 0.0- 0.00000 5.500 5.50000 0.000 000200 000 0200 0500. 0.0 5.0005 5.00 0.0005 0.00 00000000 0500. 0.00 0.500000 0.055 0.500000 0.005 0000000 0003 mocmHHm> UMWWM mU£MHHm> “WWW.” whomwubu whoomnm mocmnmmmwo muw>0uom cam Hmucmfiwuwmxm cam Houucou BUflfiZH ma mnmdfi EZHmmomm dad 3OQ UZH30mm mBADmmm GOSSHB 252 5000.0- 0.00- 0.00000 5.050 0.00000 0.000 0000000 0000.- 0.0- 0.5500 0.50 0.0550 0.00 00002 0002 0000.0- 5.0- 5.00550 0.50 0.00000 0.00 0000000000 0000.5- 0.0- 0.0050 0.00 5.00000 0.00 0000000-0200 0000.0- 5.0- 0.0500 0.55 0.00000 0.00 000000 0000.- 0.5- 0.5005 0.00 5.0055 0.00 00>00e 0003002 5000.- 0.0- 0.05000 5.000 5.50000 0.000 000200 000 0200 0500.0 0.00 0.0500 5.00 0.0005 0.00 00000020 0000.0 5.05 0.505050 0.005 0.500000 0.005 0000000 0002 0000000> 0MWMM 0000000> 0MWMM 00000000 0000010 mocmumMMHQ 000>wuo< cam Hmpcmaflnmmxm :00 0000200 000020 02000000 000 2000 0203020 0000000 002200 on mnmde 253 a reduction in semi-leisure activity becomes much stronger with a decrease of 9.0 minutes for a total of 90.3 minutes per day. Other changes indicated are a reduction in leisure activity of 10.6 minutes to a total of 128.2 minutes per day. Other changes appear in a reduction of eating time, a slight reduction in television time, and interestingly an increase in the amount of work-related time of 20.7 minutes to a total of 280.5 minutes. It appears that for shopping to show any impact, it is necessary for the activity event to be at least 50 percent longer. The primary impact at this high level seems to be in the reduction of semi- leisure and leisure activity, with an increase in the amount of work-related time. Impact of Combined Experimental Conditions The strategy for the combined experimental con- ditions is much the same as for the various experimental conditions considered singly. The combined experiments represent the most realistic view of the potential impact of BCN technology because it is most reasonable to assume that the technology would impact in all of the experimental areas. The first of the combined experiments is a combination Of all of the low treatment conditions, representing an estimate of the minimal overall impact of BCN technology on marketing. 254 The combined low treatment conditions for trans- portation, television, and all shopping are shown in Table 21. The chi-square for this experiment is 1.26, again with a nonchance probability of less than .005. While there is a 1.6 minute increase in shopping to 36.0 minutes, and 2.2 minute decrease in nonwork travel, and a 2.9 minute increase in television viewing to 91.4 minutes, overall there is very little impact, except for a very slight reduction in semi-leisure activity of 6.8 minutes. At the low treatment level, the main effect seems to be to trade off decreases in nonwork travel time with increases in shopping and tele- vision time. The combined high treatment conditions shown in Table 22 exhibit by far the greatest impact of any of the experiments. This experiment has a chi-square of 17.14, which translates to a nonchance probability of nearly .975. Shopping time is increased 11.3 minutes to a daily total of 45.9 minutes, television viewing time is increased 23.5 minutes for a daily total of 112.0 minutes, and nonwork travel is reduced a total of 18.3 minutes for a daily total of 32.5 minutes. It is extremely interesting to note that only slight changes exist in any other category, including a 5.9 minute reduction in leisure and a 5.5 minute reduction in home and family activity. It appears 255 hmmm.| m.v| «.mmmma o.vma m.mmmmH m.mma wusmwmq beam.| m.| o.mmam n.mw v.mn~m m.¢v mwcmz mmmz mmmm. m.m H.Hmama v.Hm o.ommva m.mm cowmfl>mame mvmv.au m.m| m.mmnm m.mm n.mmmoa m.mm mHSmwmAnHEmm ovmm.n m.m| H.5mooa N.Hm o.mmmoa H.vm mcfiumm momm.n m.~| m.mmv~ . m.mv n.mm>~ m.om Hm>mue xMOBQOZ mama. m. m.vmoma ~.Nma n.5mmma m.HmH mawemm cam mEom Nome. a.H h.vmv~ o.mm m.HmHm m.¢m mnammoam amen. m.m «.mmnvaa o.mmm m.nmmoaa m.mmm omumamm xuoz mocmwum> ummmm mocmflum> uwmmm muoomnu mocmumMMHo . . mmwmwwww cam Hmucmfiwnmmxm cam Houucou . BudmZH EZHmmomm Aflfi 30A QZfl ZOHmH>mHmB 3OQ .ZOHfidfimommz¢ma 30A UZHBOmm mBADmmm QQZEHB Hm mnmda 256 homo.flu m.mu m.~omma m.~ma m.mmomH m.mma musmamq mane. m. n.0mmm a.¢q v.os~m m.¢q magma mmmz vamm.m m.mm a.moam~ o.~HH 0.0mmvfl m.mm conH>mHma mmvm.: o.m- s.aflmoa m.mm s.mmmoa m.mm wusmamqnaEmm avmm.u o.m- n.5mHoH H.Hm o.mmmoa H.vm acflumm mmmm.m- m.mH- o.vsHH m.~m s.mma~ m.om Hm>mua xuozcoz ommm.- m.m- n.5aamfl m.mma a.smmma m.HmH mHaEmm was meow mmmm.¢ m.HH m.maam m.mv a.HmH~ m.¢m maammoam Heme. m.H o.maH~HH H.Hm~ $.5mmoafl m.mm~ umpmHmm xuoz mUQMflHm> pmwwm mocmanm> “WWW” muommumo muoomuN mocmuwmmwo wufi>fiuom cam HmquEwummxm cum Houucou BudeH OZHmmomm AAfi mem Q24 NN mnmdfi ZOHmH>MAmB mUHm .ZOHfidamommZ¢mE mUHm GZHSOSW mequmm DOZZHB 257‘ that if all of the effects suggested in Chapter II are included in the experiment, that there is a tendency for them to cancel each other out. The main impact of BCN technology then appears to be the trade off of nonwork travel for shopping and television viewing time, with little impact on other activities. The single most important factor here seems to be television, which is partly due to the length of time, relatively speaking, and the time of day, primarily the evening hours, that the activity is engaged in. Table 23 and Table 24 show experiments holding the other treatments at the high level, transportation and all shopping, but manipulating the television condition. Table 23 shows an experiment with transportation and all shopping at the high level, and television at the nontreatment level. The chi-square for this experiment is 12.62, which con- verts to a nonchance probability of nearly .90. The increase in television viewing time drOps dramatically, showing only a 2.2 minute increase. This experiment still shows large changes in shopping with a 15.8 minute increase, and in nonwork travel with a 15.7 minute decrease. The only other change of interest is a slight 5.2 minute decrease in leisure time. Table 24 shows the same experiment as shown in Table 23, except television viewing time is reduced to the low treatment condition. The chi-square here is 258 momm.u «.m- m.mHHmH m.mm~ m.mamma m.mma magmamq «mos. ~.H m.msmm “.mq v.os~m m.vq meow: mmmz «Has. m.m m.~mmafl n.0m 0.0mmvfl m.mm coamfl>mflme mmam. o.H m.mm~HH m.OOH h.mmmoa m.mm musmamq-HEmm ¢~o>.- ~.m- o.o~mm m.om o.mmmOH H.am maaumm amas.su s.mH- ~.smma H.mm a.mms~ m.om Hm>mue xuozaoz made. a. m.mmama q.ama s.nmmmfl m.HmH »HHEmm cam mace memo.m m.mH m.moov v.om m.HmH~ 6.6m maammoam axed. m.H o.vmvmaa a.ao~ 6.56moaa m.mm~ umpmflmm ego: mocmwnm> umwwm mommaum> uwwwm whommumu muoomlm mocmHmMMflo >ua>fluo¢ cam Hmucmfiwuwmxm cum Houucou BU4ASH OZHmmomm and mem DZ< ZOHmH>MAmB 02 mm mamflfi .ZOHBdfimommz¢ma mUHm GZHSOmm mBADmmm DQEZHB 259 «mov.au 5.5: >.mavva H.Hma m.mmmma m.mma whamflmq mmmH.I v.| m.am~m H.vv v.onmm m.vv mwcmz moms Hmam. m.v v.mmmma o.mm o.ommva m.mm coflmw>mam9 maam.al o.m| m.onmm m.mm b.mmmoa m.mm madmwmAIHEmm hmom.| H.v| m.¢vnm o.om o.mmmoa H.vm mcflumm vvoa.m| m.wH| v.amma m.vm n.mmhm m.om Hm>mne xuozcoz 0665.- m.v- v.5HmmH o.sma 5.5mmma m.HmH madamm can mace mamm.m m.ma «.maflv «.mv m.HmH~ m.qm mcammoam mmom. n.ma m.mmhbaa m.m>m m.bmmoaa m.mmm pmumamm xuoz mocmflum> uwwwm mOQMflHm> uwmmm muoomlN mocmumMMflo whommumo cam Hmucmeummxm cam Honusoo auw>au04 BudeH GZHmmomm AA4 30H: 02¢ ZOHmH>mAMB 309 vm mamfla .ZOHfiflfimommz¢mB mem UZHSOmm mBADmmm QOSSHB 260 12.49, again nearly a .90 nonchance probability. This experiment shows results similar to the previous experi- ment with a 13.6 minute increase in shopping time, and 16.3 minute decrease in nonwork travel time. The other changes, although representing only slight trends, include reductions in semi-leisure activity of 6.0 minutes and leisure activity of 7.7 minutes. This indicates that while shopping and nonwork travel seem to balance each other, television has some capability to compete with other activities such as semi-leisure and leisure. CHAPTER VI--NOTES 1Joan E. Jacoby and Stephen Harrison, "Multi- Variable Experimentation and Simulation Models," in Computer Simulation of Human Behavior, eds. J. Dutton and W. Starbuck TNew York: John Wiley & Sons, 1971), p. 627. 21bid. 31bid. 4George S. Fishman, "Problems in the Statistical Analysis of Simulation Experiments: The Comparison of Means and the Length of Sample Records," in Computer Simulation of Human Behavior, eds. J. Dutton and W. Star- Buck (New York: John Wiley & Sons, 1971), p. 659. 5Janet Tomlinson, N. Bullock, P. Dickens, P. Steadman, and E. Taylor, "A Model of Students' Daily Activity Patterns," Environment and Planning 5 (1973): 265. 6Ibid. 7Quinn McNemar, Psychological Statistics, 3rd ed. (New York: John Wiley & Sons, 1962), p. 220. 8Summarygof United States Time Use Survey, Survey Research Center, Institute fBr Social Research, Uni— versity of Michigan, Ann Arbor, Michigan, 1966. 261 CHAPTER VII SUMMARY AND CONCLUSIONS Summary The basic purpose of this dissertation is to assess the potential impact of broadband communication network technology on consumer marketing communication using a computer simulation model of human time allocations. BCN technology represents the combination of interactive cable television, computer and information processing tech- nology, and communication satellites. The computer simu- lation model is used in an experimental mode to determine the impact of a hypothetical BCN system on the allocation of time to daily human activity. An analogy can be drawn in the assessment of television from a 1935 perspective in describing the approximately ninety minutes per day the average adult spends watching television in 1965. The goal is to estimate the impact of BCN technology on con- sumer marketing communication in a similar manner. Marketing communication can be described in terms of a number of traditional marketing activities, including personal selling, advertising, packaging, point-of-purchase, 262 263 direct mail, product sampling, publicity, and public relations. Each of the traditional marketing communi- cation situations can be described in terms of communi- cation situations, such as face-to-face in the situation of personal selling and mediated in the situation of advertising. Traditional consumer marketing communication activity can be organized in terms of a diagrammatic communication model with a source, message, channel, receiver, and feedback. A typology of marketing communi- cation methods is suggested using the elements of such diagrammatic communication models. The first type of marketing communication is advertising and promotion which involves the passive consumption of messages by large numbers of consumer receivers through the mass media. This type of marketing communication for the most part does not involve any immediate feedback. The second type of marketing communication is retailing and selling which involves the active participation of both the source or seller and the consumer receiver. The communication situ- ation is often face-to-face and usually implies transpor- tation in order to accomplish the communication. The third type of marketing communication is marketing research which is described as the feedback mechanism from the consumer receiver to the marketer source. The problem with market research as a channel for feedback communication to marketers is that the communication is 264 primarily controlled by the marketer which may eliminate some of the communication by focusing on material of interest to the marketer rather than of interest to the consumer. The impact of BCN technology can be considered on each type of marketing communication. The impact of BCN technology on advertising would center around the addition of response capability and the addition of con- siderably more program material. The result would most likely be a more attractive program offering which would command a larger share of the viewer's daily time. Such an increase is indicated among traditional cable tele- vision subscribers. The impact of BCN technology on retailing and selling involves two dimensions. The first dimension is the direct trade-off by consumers of communication time for transportation time that is implied in the offering of in-home shopping services. The second dimension is an increase in the amount of time shopping which may result from the reduction of the overall cost of shopping because of in-home shopping services. This latter impact is suggested by a study of shopping behavior where retail outlets are in geographic proximity and where they are not. Another important consideration of the impact of BCN technology is the suggestion that the better edu- cated and more affluent consumer would make better use 265 of the technology, possibly serving to widen the gap between the affluent and less affluent consumer. Evi- dence for this is apparent in the studies of convenience- oriented and in-home shoppers who tend to be both more affluent and better educated. The impact of BCN technology on marketing research is more difficult to describe in terms of changes in the life-styles of consumers. BCN technology has the capacity to revolutionize data collection techniques in marketing research by using the response generated by an in-home terminal. While it can be argued that improved research techniques can result in better products and services for consumers, it is difficult to describe any direct effect on the life-styles of consumers. The means of describing these impacts more fully is found in the methodolOgy of technological forecasting. Technological forecasting is a relatively recent develop- ment among the tools of scientists and social planners. Technological forecasting originates primarily from the administration of military research and development and is divided into four methodological groups. Dialectical methods are founded on the notion that both the future and past history are the result of a sequence of conflicts. The dialectical method itself is divided into two major approaches, the scenario and the Delphi Method. The scenario is perhaps the most 266 common approach in technological forecasting and involves the expert consultation of a large volume of relevant material and literature, and then the application of intuitive judgment to describe the future. The scenario approach is much the same as that of the historian. The Delphi Method is a unique method of eliciting and refining group judgment. The salient features of the method are anonymity, controlled feedback, and group response. The basic theory behind the Delphi Method is that with repeated measurement the range of expert responses will decrease and converge toward a midrange, and that the total group response will successively move toward the true or correct prediction. Teleological methods explicitly recognize the interaction between the forecasts made and the future itself. In other words, the forecast itself plays a role in the determination of future events because the forecast becomes a goal. An example of a teleological forecast is the normative backcast or PERT analysis. In PERT, a sequential network of activities is constructed with time estimates for completion of the necessary activities. Experimental and empirical methods are data oriented technological forecasting methods. The most common example of this method of technological forecast— ing is time series analysis. Time series analysis allows accumulated past experience to guide future 267 expectation through the numerical representation of some set of empirical observations. The fundamental process involves the extrapolation of trends from a mathemati— cally described series of historical points, such as linear regression. Analytical methods involve the model oriented approach to technological forecasting. Analytical methods are generally theoretically based models trying to describe the interrelationships between variables. The distinction between empirical models and analytical models is often unclear, but is generally defined in terms of the starting point of the forecaster, whether it be in data or in a subjective view of some relation- ships that can be converted into a model. Analytical models themselves can be divided into two types, deter- ministic models which are generally mathematically sophisticated models which offer unique solutions, and stochastic models which make use of random processes and offer specific case results rather than unique solutions. Normally stochastic models are thought of in terms of computer simulation models. The technological forecasting model developed here is a computer simulation model based upon human time allocation data. The model, named TIMMOD, allows for the vicarious experimentation with various aspects of human time allocation to daily activities, including 268 hypothesized impacts of the result of the installation of a BCN system. Empirically the allocation of time to various activity categories is determined through the use of the time-budget study. The time-budget study, through the utilization of survey research tech- niques, measures the duration, frequency, and sequential order of human activities throughout the day. Generally the time-budget data are considered as dependent variables, with other socio-economic and motivational factors as the independent variables. The empirical basis for TIMMOD is the United States Time Use Survey conducted by the Survey Research Center of the University of Michigan in 1965. The survey, consisting of a national sample drawn from forty-four metropolitan areas, was part of the Multinational Com- parative Time-Budget Project. The study describes the average daily activity patterns for a sample of 1,244 employed adults. The suggestion for the develOpment of a simulation model based upon time-budget data, such as TIMMOD, comes from urban planning. This kind of simulation model would allow the study of changes in activity choice and structure in an urban environment without having to make changes in the physical world. Also suggested as an application of this kind of simulation model is the study of changes in location choice. 269 TIMMOD constructs time allocations to various daily activity categories from the selection of random deviates from two basic theoretical probability distri- butions. The first probability distribution describes the activity selection behavior of the hypothetical population, and the second probability distribution describes the duration of the selected activity. Both probability distributions are estimated from the data published in the United States Time Use Survey. The simulation begins with a single individual, who has a randomly selected waking time and retiring time. Accord- ing to the appropriate hour of the day, an activity is selected, followed by the computation of the appropriate activity duration. The process continues until the entire day is exhausted. When the first individual has retired, then the process starts again with a new indi- vidual. The activity selection distribution consists of a matrix of cumulative percentages by activity category and hour of the day. Uniform random numbers are drawn and compared to this matrix to select a particular activity. The activity duration is determined from a family of Erlang distributions with different para- meters to describe different shaped distributions for different activity categories. While the total time allocated to various activity categories is the TIMMOD output, the input can be one of 270 three possible variables. The first is an activity selection weight which can increase or decrease the probability of the selection of a particular activity. The second is the constant parameter associated with the Erlang distribution that determines the shape of the distribution from an extremely skewed exponential dis- tribution to a nearly symmetrical normal distribution. The third is the other Erlang distribution parameter which determines the length of the activity for the particular distributional shape. Experimentation with TIMMOD then involves the manipulation of some combination of these input para- meters with the associated comparison of the output time allocations. The experiments described here involve only the manipulation of the length of the activity parameter. The other parameters are held constant. The experimental conditions include the manipu- lation of the parameter to reduce the amount of time spent in personal travel, which represents the communi- cation for transportation trade-off aspect of the appli- cation of a BCN system. Another experimental condition involves the lengthening of the time spent viewing television, accommodating the assumed increased attrac- tiveness of the medium as a result of a BCN system. Also considered is the lengthening of the time spent shopping for both nondurable or everyday goods and the 271 time spent shopping for durable consumer goods. The experimental conditions are appropriately named transpor- tation trade-off, television viewing, marketing (or shopping) for nondurable goods, shopping for durable goods, and all shopping which encompasses all shopping behavior, i.e. for nondurable and durable goods. These experimental conditions are considered separately and in various combinations. Two levels of each experimental condition are considered to simplify the number of possible simulation runs. There is the low experimental treatment level which represents a 10 percent increase or decrease in the appropriate category, and the high level which represents a 50 percent increase or decrease in the appropriate category. In general, each of the experimental conditions demonstrated almost no impact whatsoever with the low treatment level. TranSportation trade-off time exhibited only a very slight trend to decrease time allocated to shopping activity and increase time in home and family activity when the personal transportation time Spent per trip is decreased at the high treatment level. The high television viewing time treatment has somewhat more impact than does the high transportation trade-off time treatment when considered alone. With the high level increase in television viewing time, there is a trend toward a decrease in other mass media 272 time and in leisure time. There is also a small decrease in time allocated to home and family activity. Overall, television viewing time appears to be the single most important of the experimental treatment conditions considered. Shopping represents a more complex situation. In the low and high level treatment conditions for non- durable goods shopping or marketing there is almost no change in the allocation of time to other activities. In the case of the high level durable goods shopping condition there is a very slight decrease in semi-leisure activity time which includes such activity as personal care, study, religious and organizational activities. Combining both nondurable and durable good shopping into an all shopping category, a stronger trend toward a decrease in semi-leisure time is apparent in the high level all shopping treatment condition. Other changes include a reduction in general leisure time, eating time, a slight reduction in television viewing time, and of most interest an increase in work-related time. The combined experimental condition of low transportation trade-off time, low television viewing time, and low all shopping time shows almost no overall change in the time allocated to other activities. If there is any effect, it is a trade-off of the decrease in travel time with increases in shOpping and television 273 time. The same pattern, although far more apparent in the high level transportation trade—off time, high level television viewing time and high level all shopping time experimental conditions, is apparent. In this all high combined experimental condition all of the major changes appear in the manipulated activity categories, that is nonwork travel, television viewing time, and all shopping time. The effect of the three experimental conditions seem to be cancelling each other out. There are only a few other noticeable changes, including a small reduction in the time allocated to leisure activity and home and family activity. Of the three experimental conditions, television viewing time appears to have the greatest impact. Holding transportation trade-off time and all shopping time experimental conditions at the high levels, and moving television viewing time from the low experimental level to no experimental manipulation at all demonstrates television's ability to compete with other activity such as semi-leisure and leisure activity which fill the void left by television time. Conclusions The conclusions which can be drawn from this research fall into two general categories. The first category deals with the application of the computer simulation model to the technological forecasting 274 problem. The second category deals with estimation of the potential impact of BCN technology on consumer marketing communication. The methodological conclusions are considered first. The TIMMOD experiments demonstrated impact on the human allocation of time to daily activity categories. The amount of time allocated to particular activity cate- gories changed in response to an experimental manipulation of various treatment variables. Examining the relative changes caused by a particular experimental condition alone makes a case for the validity of TIMMOD. Reducing the amount of time spent in shopping related travel, for example, reduces the amount of time spent shopping and increases the amount of time Spent in home and family activities. Increasing the amount of time spent viewing television reduces the amount of time spent participating in other mass media consumption and reduces the amount of overall leisure time. Increasing the amount of shopping time, that is making shopping a less expensive activity, reduces semi-leisure time, that is time Spent in personal care, study, religious and organizational activity, and reduces overall leisure time. Most interesting, however, is that increasing shopping time also causes an increase in the amount of work-related time. Increased shopping time is the only experimental condition to demonstrate a concomitant increase in work-related activity. TIMMOD 275 results appear to have validity in terms of the relative changes in time allocations. However, the magnitude of the changes may be another question. It is interesting that TIMMOD does not demonstrate much change as a result of small or low level change in treatment conditions. For example, the low treatment of transportation trade-off time involves a 1.6 minute reduction in the average time spent in shOpping related travel per trip, the low television treatment involves an increase of 6.9 minutes in viewing time per time watching television, the nondurable good shOpping and durable good shopping trip times are increased 3.2 minutes and 3.9 minutes respectively. At these rela- tively low levels, TIMMOD demonstrates little or no change in the allocation of time to activity categories. At the high treatment levels, indicated by a 3.9 minute reduction in the average shopping trip, a 34.6 minute increase in the average time spent watching television, a 15.7 minute increase in the time spent shopping for non- durable goods per shopping trip, and a 19.6 minute increase in the time Spent shopping for durable consumer goods, there is a marked impact on the time allocations. In other words, TIMMOD does not appear to be sensitive to small changes, requiring relatively large changes in the experimental conditions to make obvious the relative impact of the experimental condition. TIMMOD does, how- ever, demonstrate the ability to show differing impacts 276 by activity category not necessarily in proportion to the amount of time increased or decreased in the experi- mental condition. This is due largely to the fact that TIMMOD makes use of an activity selection probability matrix that changes with hour of the day, and allows for different shaped probability distributions to describe the various activity durations. Overall, TIMMOD appears to have potential as a useful tool in future planning. Given the limitations set forth for the application of the model, the experi- ments conducted using TIMMOD were successful. Within the limits imposed by the available information sources, that is time-budget study data and data describing various aspects of the marketing communication process such as shopping, TIMMOD provided a successful demon- stration of an analytical technological forecasting model. Turning to conclusions about the potential impact of BCN technology on consumer marketing communications, it appears, relatively speaking, that BCN technology will not have an impact in the same order of magnitude that broadcast television has had. Assuming that BCN tech- nology would result in a combination of benefits, includ- ing a reduction in the amount of time spent traveling, increased attractiveness of television due to increased program material diversity and the added interactive 277 capability, and an overall reduction in the cost of shopping, there appears to be no impact on other human activities except on these three. In other words, decreases in Shopping travel time are offset by increases in television viewing time and shopping time, with little apparent effect on any other activity category. Unlike television which was able to attract its average of ninety minutes per day in viewing time from other activities, BCN technology appears to have little effect on other categories. However, Semi-leisure and leisure activities appear to be the most vulnerable if there is any impact on other activity categories at all. -jBCN technology does not represent a new mass medium in the same sense as did television in 1935. Time allocated to the other mass media and other activity categories would remain much the same today3\ There is, however, the potential for the redistribution of time among the shOpping related categories already discussed, including transportation trade-off time, television view- ing time and shopping time. An obvious impact is the reduction of travel time in the range of five to thirty- five minutes per day depending upon the volume of shOpping services available on a BCN system. This travel time reduction could be translated into dollar savings, which should become more important as the price of fuel increases. There is a physical distribution problem ll .4-“ m..___. o 278 which has not been considered here, that is moving the physical goods from the retail outlet to the home. The majority of this physical distribution currently is accomplished by using the family automobile. A BCN shopping service implies some form of home delivery service, which presumably would be more efficient than the use of individual autOmobiles. 3~Another impact of a BCN system on shOpping is the rather intangible result of increased shopping time. Consumers who spend more time shopping might be expected to become more expert in making purchase decisions, and thus become more efficient shoppers. A consumer pre- sumably would examine more purchase alternatives. Add increased efficiency to the improved product and service offering resulting from improved marketing research techniques? and it is possible to forecast improvements in thequality of life as a result of the introduction of BCN system services. A more cynical view would be that the increased shopping time would result in con- sumers buying more products and services that they really do not need. In the short run, BCN technology offers a reduction in shopping related travel, which most con- sumers probably do not consider particularly important, an intangible benefit of increased Shopping time, and perhaps a more entertaining form of television. The 279 consumer benefits of a BCN system may be seen as either relatively unimportant or obscure to the consumer. This lack of obvious consumer benefit may be a contributing factor in the retarded growth of BCN systems. This does not mean, however, that the benefits of such a system always will remain obscure or unimportant. Suggestions for Future Research It is a simple matter to suggest the further refinement of computer simulation experiment methodology, and point out that related statistical analysis tech- niques are needed to improve the ability of such method- ology to serve as a technological forecasting methodology. However, this type of inquiry would provide only improve- ment in the efficiency of the computer simulation model, such as TIMMOD, without solving the fundamental problem of specifying experiment treatment conditions necessary to conduct vicarious experimentation. Similarly, it would be simple to point to the lack of time-budget data needed to develop more precise activity selection sta- tistics and activity duration statistics for different hours of the day. Also, activity selection and duration statistics could be calculated for different sub-groups within the population, such as housewives and children. While such data would be of enormous value in refining TIMMOD, it would be very expensive to collect because it would require large national surveys. While such 280 data would be useful, the resources necessary to obtain them could better be directed toward research that would solve the more immediate need of better description and specification of the experimental conditions. As is apparent from the difficulty discussed throughout this work, the most critical need for addi- tional research is in both the description of the funda— mental processes that may be involved in the application of BCN technology, and in the demonstration of the capa- bility of the BCN technology to modify these fundamental processes. Examples of such processes include seemingly elementary behavior such as shopping and viewing tele- vision. As mentioned before, there is a need to under- stand better in precise terms the role played by the family automobile in the physical distribution of goods. A personal automobile utilization study would be rela- tively easy to undertake using standard survey and diary techniques. The result of lower shopping costs also could be more precisely specified. For example, the relation- ship between reduced sh0pping costs and the time allo- cated to shopping needs description. Also the difference, if any, of product category on shopping time needs to be understood. In terms of its activity characteristics, shopping is not particularly well described in the research literature. A study of cable and noncable television subscribers should be considered to get a better under- standing of the relationship between program offering 281 and television participation. A demonstration project applying a two-way digital return system would be help- ful in understanding the dimensions of any increased viewing time that may result. In addition to studies which would better describe the fundamental processes involved, it is mandatory to develop commercially oriented demonstration projects of the BCN technology. Only through such studies will the parameters surrounding the impact of BCN technology on consumer marketing communication be understood with any degree of certainty. When these parameters are better understood and the range of treatment in terms of input conditions better specified to TIMMOD, then any number of experiments can be conducted to estimate the potential impact of BCN technology on the life-style of users of the system as indicated by the way they allocate their time. APPENDIX TIMMOD PROGRAM ,32),STOT(3,32) T- 0 T S N 5 5‘ HT RC 0.7: 5 0 NT 1 5 1* 99 1 1 1 S), 1 o ‘1 , E22 0 0 O 2 2 H33 T 0 T 1 S 0 0 It: T = v 7 ; 7930 0 o \l \I 2 ‘1 N12 6 0 G .I o 0 T1 5 .1 \I o 0 \1 o SAX \) \I F" 9 J o T o J 0 NMI ) \l 9 S 0 T3 9 6 fl: 9 F. )UTR 1 3 T H 5 A( I t 0 1 I . TRcT 1 h I R 1 CRT ( 1 G ( ) UN’A 9 9 9 p OUA T9 J 0 T J 92" 1 0 56:... E N 0 1nH 0 A 9 \a .l A 9 3P : 2 H1H N : T T 5 H 2 0 H 2 U599 K :1 T11 I T ++_ 2 T ( o O T ( 050) R E: T T T C 0 ) T 1 G T 0822 \I ‘90 HE)IOI I A G ’70 0 t. A O C A 7.113 2 0H [Hauleul‘ ( I TT 0 T M Q .1 M. U 9X1 ’1 0.... TT. 0X X X c.) 110 .9 T E S 9 T. P)IQ 1G 0 )ITOTOI I ))TP 99350 J ( o E J ( N5RS (N 0)0 IORGR R STAEH 120TG : ( ) H 9 ( 7.1!.) CL 10 o) 1.! TTT HCREO ((EA .1 T. I T. (“A 9 NR 9 PHD) 9A1ACA RAULN TT R) T l. 0 01.! T .l 1 0 03") LE 01 Z10H1HAH PIOST AA)U1 C T T C N C T 0(A2 R 05. o 0 9A DANA N1 o . "M10 0 A 0 \I 9‘ o O A o ‘9 HM.3 E. O- ETOoEovo oGoEP TTI 0 N31 J N3. T N3T J HAT‘ (THC L1TNE1E «(T-N LE : .. 99F. 99'... 1! col 990 L 9 lb 1 T. IRCQ 990 M60 0 0H6 0L ..L GLH 0F. 993 o 11.. Q 3811: o 11: Q R TAAS U .5 RoNEERoKoTo QoROTL TT(HTO=:) S (E::) T0:=) 5 OCT PNT T05K FKTHHFHRHIHE TENSS IITOH1J1JETETTJIJEESlJIJESEOPI P 0E oZ‘nunuN R(IINUCU UUTC :TE .. 1 9 ‘NG1 9U : UAN 9UUE3 9U .. U3EX ANNN:SQ:H==U0:TTUNHN0NNIAE:TE:12HTI 05IN)NN=05INNT SHIN9N RE ROOOT :P:TK:HEII=RIR2RI:ITTSTT((T:N06611JITT88‘ITNO77(IJIO GHHHSLEEHH K(H((H(((1(TT(AS(AHTT(H(T?2TT(T(SZZTTT(T22TT(TTLL OHHMELKEOG R I U NC RF RGAA 0 ONQN E ANN ONQN L ROOOTAALNIHOFTFFNFFFOFDAFUTFHIHHFNFOOOTOSOFTOONOOFOOOTOSOOA PCCCNCHSTNIHIIIIPIIIDICIIOSIDNTTITIGDDTCTCINDDTCCIGDDSCSCGC 0 0 5 6 00 5 0 0 0 5 nu 0 50 0 Q 5 0 0 1 1 1 2h 9 5 0 5 6 6 0 88 7 7 7 1 1 1 1 1 11 1 1 2 2 2 2 3 2? Z 2 2 3 5 0 5 0 5 0 5 0 5 0 5 0 1 1 2 2 3 3 Q 14 5 5 6 282 283 SUBROUT XNE SETUP ’ 9 = T c I 1 N z O 3 9 9 x 9 5 O u 9 N T T 5 O I O 1 1 O T 9 T X s . U ’91 T9 3 a [N 09 m 1 OT T2 R XI 53 R T 1N N 9 P s R o 3 I. N T TTN C. NT 9 O 9 N1 5 "~09 I T X T T J NT 3 M. S H "We... 9 lo 9 . T 99 I 9 15 L TPL O $.99T T ’1 1 INS T EZZI C .9 o 9 N NT 5 N330 I 5 So I Y N 1 9 9P N 9 R 81. NNT, 9 9 1 T39 9 9 2 E510 001.9 T .9 .9 0 .9 0 N12!- 0 9 9 TE 1.1" c \l o . 2 N 2 899.9 s. K T 9 N9. TUTIS I 9 Z 1 9 1 9 ix,- E 0 9| 9 T. 9 g! 9 N 99 1 9 9 1 9 1 WNTT s 9 N 9 N . ITO Tot—rib; 9 9 5 9 1 1 1 1 1 .91 RN J .90 I J U 0 .9 RN: XLLI 19 .9 s 9 1 1 I n a a TI... ”TU 9 0?. 9 9 S 1 T IonT 55E 6! X N 1 x J J J J 167.3 )Is 5 T 000 19 I 1 I I 9 PT ass-EFN 3 n I h. J 9 9 9 9 N199 C". n K: O O 0 9|. P 1 [‘99 J EN N v 9' 9 J 9 9 T 99 ‘9 \I 9 75 D3P 9 N 9 009 11 9 T 9JJ’ 1 T559. 77191 3 N 9 ‘9 9 1 1 1 1 1912 £999 1 666 I! 9 U J99J. o X UTU TTT-h o 9 I. 1 T. 9 9 9 9 2...... :0)? N .. III KK K P 9919. o 1 9590 TI. 7 S J 9 9 J J J J KKKK $223 9 K 11,1 99 9 N 1110 o N "19 F'VL.‘ F E 9 J J 1 1| 9| 1 9999 113 9 s 9 123945 1 J T (Xvi-.0 0 T 0 9.. 1V 1 1 9‘ 9. T X T 1.9)) 1(0 E 1 99999 11 ( I 9.1!! 1 I o 95'. STT 20 0 H n... T N I "u I KKKK )ane, n ‘ To!!! 9' T ”3'... i n . w 2 991 ecu}. 1 T m N Q R a R (19“ 3512/, 1 9 1111 U 9 TN 9 P 2 55V XIICTI 9 N M P T T T T TTTT 3159.300 T 1 III II (I. P NIIT 1 9 I T N11 59 . NIX X 9 R N N I I I I CCCC II 901 o 9 TTNSC QTTTTT TT N 9 TNNIS 9 \9 9 EU T9 99 9° .5 5 s (I T T N N TN TN PPPP TSNTPTOI N1 NI NIIIII TUU T 1 TI. TIPPNX T T T T 1 SN. C/ZINNI 9 9 N NP TI TI TI TI GP G? PPPP ZNCOI .09 T moon c, c N 883‘. c cos c 1 0| .N:I.1|o.l’u z m 9| 6‘ CI ll CI ‘1 I1 9999 51920 :6 I PPP IN I“! o N ’1) \- oI I 090 2 INT T 9019059999! 1 "9 ‘9 I9 9 I9 N9 N9 NKNN ITCIP32 11FUNO N: .. .. .. : N1 N2) 2N0 ZNJJJOJ ZN 3N000 3 N20 C 1 .0 99X111I 9 1 N N1 N1 N1 N1 N1 1 1 ((19.0 ECO II TFORIP 1.91 II 99K 91) 99999 9 99 99:3: 9.0 999. P NXTXENNNN II 9 99 99 99 99 99 19 19 9999p: NIISNNTN 651N911 1123.3) 111 11 9 11I 1111"... 11 11111 1 o 9 1p P 19831-911196 61.1293 51° 21. ~11 211 6:1 7.31 7777‘. LNTOO 00(9): 9.. z 9 9 9 9 It: 9“: .. er; : :WWH.:(1(U1E0WSTWS fiJJJWE:00 s .. : E._r.((N( ((111 1090:0E0:0 0:050:0503 05010 010E0000 T 1 S . 0.11 111110122 KJ... UTJ) JIXXSXU . [I I 9 9 9 U9... .. [J9U9U 9 222120213 313U213U113U2 3W3 3U2222 UNNNSSPIOTTT1 1N (((1N 11N TNN TNN I1X1N0N IN TTTNN ))N TNTNTT TITTTTT TT N N N N N 0 N 0 N :N OOOONN ..III 0 [SNNNNNTO T12U115~¢1107RR9 2R1...19R1051991110111051111IIXINIIIIII IITTTT1T5T1T6T1T0T1T1TTT0T1T1TITTTT0R RNNNFIIINNNNOZDTZIIIIIT3DOTSBPTT33NTT33TTTNTTNT3TT~ITTTTT711T00TTTTNN5NlNNNNNN NNNNNSNTNSNTNSNTNfiNTN6NT N1NTN1NTNNNNEU BHNNNNTTUNRRI IN RRRRRN IIN NNN UNN IIIUI NUN IN IOONN 3N CNCNRR RPR RNRN RMTTT TN! TN! TNT 1N1 1N1 TNT INITITTD SCCCOOOOXFFFROR CDPPPPPCO CODICCOOICCOO IPXI CXCDICDDTTSCCDTSCOOPC C Fun-F9 FFFF F? P FF P OWCWONC DWC CPD C CPO C P TR... . '12 ' s . 21 N5 7 . 9 so ' 5 001 z 3~56‘ w 92 o s 6 o 1 o . 111 2 2 3 33 33 3 3 3 It 7 6 000 0 000000 E 09 5 5 5 6 6 0 1 22 2 222293 5 33 1 1 E N T T U 0 R 8 w 5 o 5 o 5 o s o s o 5 o 5 . s . s o s . s o 5 1 1 Z Z 3 3 I I 5 5 6 6 7 7 0 I 9 9 0 0 1 1 1 1 1 1 284 SURROUTINE REPORT 9 I 5 ’ I I \9 9 U 9 9 2 X 9 3 5 9 V 9 999’ 999999999999999 F9 I R 3 9 U I 0 1 2 L 5 EC 5 G T 3 E T NN 1 E O 9 V NE II 9 T T 2 DV- I N E To 9 I s 1 EL 8 E? S 0 VU 99D 9 R c T 9 9. TI H ST 8 1 F31 I 99 I E 09 E H" 0R U U S T IX9 9 B T T2 8 0I 155 CL L E ISLE 93X X N 0 $3 I 1F N 9HN HA C N SRIL $92 3 U T9 N9 L 9HHHHOHHHHHHHHHHHHHHH L99 9 N SI 13 9 00000J000000000000000 IT9 2 N9 51 9 L1111 111111111111111 UNE 9 Y X "T 2 E9999.N999999999999999 DEC 79“ R7 0 3 R 0 ITIGY NG. ISROR. ICN FIG 09 T 1 0 I IOTNPRONYOZEEOUR. VRI (’5 F9 NT 0 KN SEN-EIDESIUTIIVDTE T51 29: 9E 99 P RIGIR KKKNHRTUOGVNTLH DP? .ST .0 S99 5 O NVUoPROUTEATAIOOUUT N91 XTA Xh E22 9 NEIES10IOIOPESRHNCOCO 19V INC II H33 9 HTLIHNNCL X9 90 11 I992 HOIEEU 99999999999 D69ZPT 6 T30 3 0HETL1999913579135791 51X 910 9A N12 1 [357911111222223 T96 QCT 59 S 9T1 U IHHHH9HHHHHHHHHHHHHHH A99 FIS9L9 E 12 S QIX D 900002000000000000000 9L 9 9TNII9 N XX E BNI T9 911113111111111111111 IUET 9Q 9UX I .- 1 1 N UTR 90' 999999999999999999999’ ."HN V‘ANYDS T 99 1 1 I D9T999 1 1 TTT? 3PIOV9 X QC. 1 1 T N9I222 : : S N Y PSTP 9 0199 : 55 X 92N331 J J D E L T 0 9S 2LSNUTI 5 NH 0 0 I D3P11B 9 I9 N 0 EE IST P 99 9LTNNN9 E 11 SS T9T9 0 E99RTI 9 L 1) BA R VR VTN§ E9X9 TINUIEZ H TT EE T I 9 E09ICLI 2 E 02 OR I II 19300 Q90X F 05 Do 1 X5 NW 0101 0 $22VP30 9 V E9 JPSGEQCNS TOHAE 095 (RP LS5 T 993555 11 GTGT 006 1135399 1 EN19 55 RT OR CONNE90199 ZOIOL 1 )S 00 :5. TT C m 999 9X1 9990 = D :00 .900 L1... 05.1.1.9 01F9 XLrCOIC.F T 9 9 HM XS 9P9 009 9109909 I TUSIJTGONCEINP IGSTH 905 1N N1 09 001111 II 0T08 001 T5R52290I9HHHH9HHHHHHHHHHHHHHHHRN9 E95LTNRI0 T0 22TTTT 99 9I9I 119 R9TS33960900009000000000000000001?9HIoIQIOT9 So 11XXSS 11 0909 992 O0I9111/9J1111J1111111111111111PT001I0TITF90 N0T99 :=:::= 99 9191 991 P3N9RTBR099999999999999999999992. FNT'10P 11 99011 999999 33 0909 191T E1I2ICII91 SST. NHI RNII LETN R9 9 9 GIT9 9920 $300 9 9 111111211 2€2E2 2110 RH 3VPTV61GK1I1DLIC DUOVPSIVRIT. I 99N29 999922 E1EN22 9999+93TT 3 91 91 3TTT IT1TTSSSBNPENLNEP SLSSESKIIETH919XXI9911XX9300 HT9911 TTTTTTooo 99T9T 9CCT ERCQ 9/IIOL EOVPETIIRLNOCTTSGXHX11OXXHH11X99900 IOS1TT CCCCCC1TT 11010 1P9: NIISNNNRBLPH .RBCIOHEHEEEEOOCNE111133D6111330100 9 9 0TTE:00 IIIIII :Ts =1T1T :TS9 IPNTOOOIISP.IEIERHOPCLPTNBSIERN11111 1111111:::00£9NSHITTENNNNNN1=:EITTTSEI::1 T 111VT10NHHLSTSH E T99 : :Uo: :1 TSU 9 9 99 9 9 99U c:c:u 99 9 UNNNSSSTS1HOET1 9oooooooooooTTTTTSTTTTTTT 1199N:SST0+9N221313011N1P9P9N2112 OOOONNN SNSO19999020680206802IIIIIOIIIIIII0111112555012111111111111T15111111 RHHH .CEWIII I2068111112222233HHHHHHHHHHHN HSTT 119-X" H11XXTTTTTTT1TTT11111T1TTT BHHHMN TTTHHHHTHHHHHHHHHHHHHHHHRRRRRTRRRRRRR CCODN:11 ::N000000 CCN 0 ON CCU UDOOIIIIIIOOOOI0000000000000000000009OOOOOOOOPPPPOITTFO120TTTTTTOPPOOFPFPOOPPT SCCCDDDDDD111101111111111111111FFFFF FFFFFFFCTSTSCXXSIDXXCTSTTSSDTSCDITISCDTST 1230 123056789.123056 . 00123 0012305 0 0 0 1 00000 0000000 5 0 1 1 52222 2000000 1 1 1 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 1 1 2 2 3 3 0 0 5 5 6 5 7 7 2E35 SUBROUTINE REPORT .il‘ :00 \lpp 2,1)"2))I(XTINES"Z) 2,1)"2))/(STIHES"Z) ELlKgI),K=1,?),7’07(1,I),TT0T(291).TVARII),IPCY(I)g TEEEE c 0H 0H A’I)! NI'IIT 9 ,x 1.) 17.-‘2‘ 9 :(‘Ctl 01023 = “ NTIHES 1,32 ,(LAB LABEL(Kg!),K=1,2),STOT(1,I’,ST0T(2,I’,SVARCI),SPCT(I), NSTOT 32 9‘ {$517.31. =EOOOUOID €000.10 9:..U 0,0)U22nfl22 2’U“22 Z’U 7.))" 3TITIN 0 0 8888 AAAA hTTTY .3858 1: = 2 .— : \l ‘1’, 9..va , 3 K K 3 9’ ZKZK (‘(( TQTR OAnA TVTV YTSS .§§§ )’,, 123“ , ’ , , 2222 ‘((‘ 9633 “AAA (OT-TIT 05558 S: = = : =‘I))) ”"U ) ‘- K K ’3 ’3 ZKZK TPTR OAOA TVYIV TTSS 9‘09 )),, 123“ , , ’ , 3333 (((( 9888 2AAAA 1TTT7 £553 7 = .. .. : .- 3"), ’9’, U \I \l K ( 0" ’3 1.K2( (\“II ToTQ GAGA leTV .l? 33 O... ,,‘I, 17.3“ 0. ’ ' ’ “tutu“ (((( “8838 «AAA AA .TTTT 35553 1: = : = = )3), 3 ) c, 5 131‘! 99 ’5 2121 (((( T185 0... 3”) 133.4 ’ ’ , , (0666 (((l‘ TRYR78888 TT§.SIOQ.SSS .. 2 = :1: = : = \I) Q’) = \I)” EK123QEK123QEK123QEK123951239K123“ ' 9 9 ,U 9 o. 9 Q. IN01111N12222N23333N3hhth555566666N777758888N69999N 9,99% ‘l \r K K ’3 9" 2anK (1“- TRTQ OAOA TV? V .lf-Q Q O O O O \l \t n a ‘1’), 1.111) 121114 .6 ’8 ’ 9 7 , 2121 8858 (((1‘ (((l‘ TR'R3BBBB OAOAZAAAA TVTV 11.1.1 TTSSQSSSS = 2 2 :1: : : .- 3333:)‘13’ ,',’ Z:=:: 2 3’), ”” NSTOT ’ 23 123QK123QEK123§EOO 9 I. D 6 UL.“ 3:11,!!!5‘7‘!‘ fl" 23(I'Tvl37. ‘1“l\ ‘1‘!“(‘1‘1Infl‘l‘1‘t “‘(l‘l‘ll‘l‘l‘l‘h (((‘I (1“ “((((I“ (((‘I'lfl T‘(T1(R(R.INN NNN1NOTN NN1NDY18888T18RBBT1BBBHY 188 38,1389818333' BQBBIBBBBT 188887NN ODDN A ANIIIII IPNIII IPN AAAAN AAAAN AAAAN AAAANAAAA AAAANAAAA AAAAN AAAANII TP POOFVFVOR RWRORTOR RRORSOOT TTT 00737100..- TTTO OTT TTOTTTTOT TvIT OYTTTOTTTTOOTTTT 0R . 3SCDI'ISCP D. PUP CP PPOP CUSSSSCDSSSC..CDSSSSCDESSSCSSSSUSSSSCSSSSUSSSSCU SSSSCP . . 2 3 0 0 0 1 2 3 h 5 6 1 1 7. 3 E Q h h .4 H .4 1 1 1 1 1 1 1 I. 1 1 1 0 5 0 5 0 5 0 5 0 5 0 5 o 5 o 5 5 8 9 a. 0 o 1 1 2 Z 3 3 h aw 5 5 1 1 1 1 1 1 1 1 1 1 1 1 SUPROUTINE REPORT 1,2),(37A8(x,tr,1=1,«) I ’ K , I ( E A .L 95 3| 1! 14:5 OKOE k. “U 0 NH TSTIR N1NTU I INT WORD...» DPCR 0 5 1 0 6 1 END 2(35 SUBROUTINE REPORT :11 :00 ‘10-? 9::U 0,0)U22“22 2’U1422 2‘1U ’33N3TITIIN 2,I)"2))/(XTIHES"Z) 2gI)"2))/(STIHES"2) 313‘ g(LABEL(K,I),K=1,?),7*0T(1,II,TTOT(2,I),TVARII),TPCT(I)g 1, 32 ,NSYOT 1,32 1 z 3 o 14 1111 (((‘ 3888 A AAA .ITTT. is SSS 1 2 ’ 2), YIVTV 'ITSQ. OOOO )2), 123“ ’ I. O. 9 2222 (((‘ 8888 AAAA (OT-77" 8555 S: : = : —- ,,)) 9’9'U \l \l K K I.) ,3 2K2K ((ll( TRTR OAOA TVTV 'IQIQ.S 90‘. )))’ 123B , ’ ’ 2 3333 (11“ 9888 2A AAA 13.!qu £555 7 : .. : : = ‘1’), 1.. 2 2 = = )3), 8,980.99, II ‘I K K ’1! 3 2K7~K ((11 YRTO OAOA TVo'v T355 0 ’§. ‘1 \I C; 5 ‘1’), 1’ 1, 123“ 9&1 it) I. 9 I. Q. 2121 6666 (1‘1! ((11 TRTR78888 0AOA1AAAA TVTV ’TfuolT 715565585 : .. : :1: : = .— 2) )‘I : ’)‘1) 3 3 K K I, ’3 ZK’LK (1“ TRTQ OAOA TV? V TYC Q 0 O O 6. \l \l .0 8 1)), 1‘1, 12‘“ )8 ,8 9 I. 9 ’ 2121 8888 11“ (((ll TRTR38833 OAOAZAAAA TVTV .TTTT TTSSQSSSS .. : 2 :1: z = .— ’,\l) .—)333 18888 3AAAA 'vlol'lvl “$588 2 : = : : 2 ‘1’), IN01111N2222N23333N3QQ ““NSSSS“ 8856N777758888N89999N {2131513T T-T'. 23(I'T33Y It'll“ (11181111381111! “(1 (111111“ (11111111 “((11114 ((1117...- fu((.l1l‘R(R.lNN NNN 1NOTN NN1NUV1B 888'1898?18888vl 188 BBT8898189883 833818888T 188883NN OODN A ANIIIII IPNIII IPN AAAAN AAAAN AAAAN AAAANAAAA AAAANAAAA AAAAN AAAANII TPPOOFVFVORWRWWORTORRPORSOOTTTTOOTTTTOOTTTTOOTTTTOTTTTOTTYTOTTTTOTTTYOOTTTTORA §SCDI1ISCP P DP CPPPDP COSSSSCUC.SSC.~CDSSC.SC9.555055550SSSSCSSSSDSC.SSCD SSSSCP . . 2 3 0 0 0 1 2 3 Q 5 6 1 1 2 3 h h h h h k A 1 1 1 1 1 1 1 1 1 1 1 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 8 8 9 q 0 0 1 1 2 2 3 3 h h 5 5 1 1 1 1 1 1 1 1 1 1 1 1 SUQROUTINE REPORT 1,2),(31A8(x,t».1=1,~) SLA9(I,K),I 19 PRINY “05 00 150 K=1 150 CONTYNUE PRINT no ‘9 RETURN END 160 286 DRANU FUNCTION UNFRH FUNCTION NSTOT ,32)9STOT(3,32) HS 1(5 .t O. 0. C—9’ 522 I13 ) ,Vklt Q.)I0 H: §RS)U A fSDN (“A 9:..R H381E¥ ERLNG FUNCTION 5 NSTOT (3,32).STOT(3,32) NT 3 ( ,1?) D 3 p EQLNG 9 X ( 0 FOR ( a A 9 1 L H 3 ) 1 6H K 10 6 '9 I. I. r. 5 o: ’ )7 '5 R (, , G) \I 07. 3 I.- Q. 9 AJ J : RGEGRCFbR 2" O N‘NU‘NU NzRFREFREN 0 DP RET. ERIER E 2 756“ 15 20 SELECTED BIBLIOGRAPHY SELECTED BIBLIOGRAPHY Abelson, Robert P. "Simulation of Social Behavior." In The Handbook of Social Psychology, pp. 275- 356. Edited by G. Lindzey and E. Aronson. Vol. II. Reading, Mass.: Addison-Wesley, 1968. Agostino, Donald Edward. "A Comparison of Television Channel Use Between Cable Subscribers and Broad- cast Viewers in Selected Markets." Ph.D. disser- tation, Ohio University, 1974. Anderson, W. Thomas, Jr. "Identifying the Convenience- Oriented Consumer." Journal of Marketing Research 8 (May 1971): 179-83. Andreasen, Alan R. "Leisure, Mobility, and Life-Style Patterns." In Changing Marketing Systems, pp. 55-62. Edited by R. Moyer. Chicago: American Marketing Association, 1967. Arndt, Johan. "Temporal Lags in Comparative Retailing." Journal of Marketing 36 (October 1972): 40-45. Ayres, Robert V. Technological Forecasting and Long- Range Planning. New York: McGraw-Hill Book Company, 1969. Baer, Walter 8. Cable Television in the United States-- Revolution or Evolution? Santa Monica: RanHE Corporation, 1974. . Interactive Television: Prospects for Two-Way Services on Cable. Santa Monica: Rand Cor- poration, 1971. , and Pilnick, Carl. Pay Television at the Crossroads. Santa Monica: Rand Corporatian, 1974. 287 288 Baldwin, Thomas F.; Greenberg, Bradley S.; and Muth, Thomas A. Design Study for Urban Telecommuni- cation Egperimentg. Department of Telecommuni- cation, Michigan State University, East Lansing, Michigan, 1974. : ; and . Experimental Appli- cations of Two-Way Cable Communications in Urban Administration and Social Service Delivery. Department of Telecommunication, Michigan State University, East Lansing, Michigan, 1975. Baran, Paul. "Broad-Band Interactive Communication Services to the Home: Part I-—Potential Market Demand." IEEE Transactions on Communications COM-23 (January 1975): 5—14. . "Broad-Band Interactive Communication Services to the Home: Part II--Impasse." IEEE Trans- actions on Communications COM-23 (January 1975): 178-84. . "Some Changes in Information Technology Affecting Marketing in the Year 2000." In Changing Marketing Systems, pp. 76-87. Edited by R. Moyer. Chicago: American Marketing Association, 1967. Barnett, H. J., and Greenberg, E. "On the Economics of Wired City Television." American Economic Review 58 (June 1968): 503—08. Basic, E. Martin. "Development and Application of a Gamma-Based Inventory Management Theory." Ph.D. dissertation, Michigan State University, 1965. Bavelas, Alex. "A Mathematical Model for Group Structures." Applied Anthropology 7 (Spring 1948): 16-30. Becker, Dietrich, and Willibald, Gunther E. "Classifi- cation and Assessment of Telecommunication Ser- vices in Broad-Band Networks." IEEE Trans- actions on Communications COM-23 (January 1975): 63-69. Becker, Gary S. "A Theory of the Allocation of Time." Economic Journal 75 (September 1965): 493-517. Bell, C. Gordon. "Some Network-Space Dimensions." Ekistics 30 (October 1970): 270-72. 289 Ben, Constantine; Bouchard, Richard J.; and Sweet, Clyde E., Jr. "An Evaluation of Simplified Pro- cedures for Determining Travel Patterns in a Small Urban Area." In Computer Simulation of Human Behavior, pp. 400-30. Edited by J. Dutton and W. Starbuck. New York: John Wiley & Sons, 1971. Block, Martin. Two-Way Advertising: An Application of Broadband Communication Network Technology. Department of Advertising, Michigan State Uni- versity, East Lansing, Michigan, 1975. Bogart, Leo. "As Media Change, How Will Advertising?" Journal of Advertising Research 13 (October 1973): 25-32. Brail, Richard K. "Activity System Investigations: Strategy for Model Design." Ph.D. dissertation, University of North Carolina, 1969. , and Chapin, F. Stuart, Jr. "Activity Patterns of Urban Residents." *Environment and Behavior 5 (June 1973): 163-90. Branscomb, Anne W. "The Cable Fable: Will It Come True?" Journal of Communication 25 (Winter 1975): 44-56. Bright, James S., and Schoeman, Milton E. F., eds. A Guide to Practical Technological Forecasting. Englewood Cliffs, N.J.: Prentice-Hall, Inc., 1973. Broadcasting Cable Sourcebook 1975. Bucklin, Louis P. "Testing Propensities to ShOp." Journal of Marketing 30 (January 1966): 22-26. . "Trade Area Boundaries: Some Issues in Theory and Methodology." Journal of Marketing Research 8 (February 1971): 30-37. Callais, Richard T. Subscriber Response System: Progress Report. Los Angeles: SRS Division of THETA-COM, 1973 . Carne, E. Bryan. "Telecommunications: Its Impact on Business." Harvard Business Review 50 (July- August 1972): 125-33. 290 Chapin, F. Stuart, Jr. Human Activity Patterns in the City. New York: John Wiley & Sons, 1974. Chorley, Richard J., and Peter Haggett. Socio-Economic Models in Geography. London: Metheun & Co., 1967. Coleman, James S. Introduction to Mathematical Sociology. New York: Free Press of Glencoe, l964. Communication Technology for Urban Improvement. Report to the Department of Housing and Urban Develop- ment. Washington, D.C.: National Academy of Engineering, 1971. Coulson, John S. "Research Revolution at the Checkout Counter." Speech delivered to the 20th ARE Annual Conference, New York, 1974. Cowan, Peter. "Moving Information Instead of Mass; Transportation versus Communication." In Communication Technology and Social Policy, pp. 339-52. Edited by G. Gerbner, L. Gross, and W. Melody. New York: John Wiley & Sons, 1973. Cox, Edwin E., and Giese, Paul E. "Now It's the 'Less- Check Society'." Harvard Business Review 50 (November-December 1972): 6-20. DeSerpa, A. C. "A Theory of the Economics of Time." The Economic Journal 81 (December 1971): 828-46. Deutsch, Karl W. "On Communication in the Social Sciences." Public Opinion Quarterly 16 (Fall 1952): 356-80. Diamond, Steven L.; Ward, Scott; and Faber, Ronald. Consumer Problems and Consumerism: Analysis of Calls to a Consumer Hot Line. Cambridge: Marketing Science Institute, 1974. Dordick, H. S.; Chesler, L. G.; Firstman, S. 1.; and Bretz, R. Telecommunication in Urban Develop- ment. Santa Monica: Rand Corporation, 1969. Dordick, Herbert S., and Hanneman, Gerhard J. Utili- zation of Broadband Communication Network Tech- nology for Proprietary Market Research. Uni- versity of Southern California, Los Angeles, California, 1975. 291 Dunn, Edgar S. "A Flow Network Image of Urban Structures." Urban Studies 7 (October 1970): 239-58. . Social Information Processing and Statistical Systems--Change and Reform. New York: John Wiley & Sons, 1974. Dutton, John M., and Briggs, Warren G. "Simulation Model Construction." In Computer Simulation of Human Behavior, pp. 103-26. Edited by J. Dutton and W. Starbuck. New York: John Wiley & Sons, 1971. Fan, Liang-Shing. "Leisure and Time Elements in the Consumer Behavior." Southern Economic Journal 38 (April 1972): 478-84. Farace, Richard V.; Richards, William D.; Monge, Peter R.; and Jacobson, Eugene. "Analysis of Human Communi- cation Networks in Large Social Systems." East Lansing, Michigan State University, 1973. (Mimeographed.) Fishman, George S. "Problems in the Statistical Analysis of Simulation Experiments: The Comparison of Means and the Length of Sample Records." In Computer Simulation of Human Behavior, pp. 659-67. Edited by J. Dutton and W. Starbuck. New York: John Wiley & Sons, 1971. Foote, Nelson N. "Methods for Study of Meaning in Use of Time." In Aging and Leisure: A Research Per- spective into the Meaningful Use of Time, pp. 155-76. Edited by R. Kleemeier. New York: Oxford University Press, 1961. . "The Time Dimension and Consumer Behavior." In On Knowing the Consumer, pp. 38-46. Edited by J. Newman. New York: John Wiley & Sons, 1966. Forrester, Jay W. Urban Dynamics. Cambridge: The M.I.T. Press, 1969. FORTRAN Extended Reference Manual. Control Data Cor- poration: Sunnyvale, California, 1973. Gabor, Dennis. "Social Control Through Communication." In Communications Technology and Social Policy, pp. 83-96. Edited by G. Gerbner, L. Gross, and W. Melody. New York: John Wiley & Sons, 1973. 292 Gensch, Dennis H. "A Computer Simulation Model for Selecting Advertising Schedules." Journal of Marketing Research 6 (May 1969): 203—11} Gillett, Peter L. "A Profile of Urban In-Home Shoppers." Journal of Marketing 34 (July 1970): 40-45. Goldberg, Victor. "Marginal Cost Pricing Investment Theory and CATV: Comment." The Journal of Law and Economics 14 (October 1971): 513-16. Goldhamer, Herbert. "The Social Effects of Communication Technology." In The Process and Effects of Mass Communication, pp. 897-951. Edited by W. Schramm and D. Roberts. Urbana: University of Illinois Press, 1971. Goldmark, Peter C. "Communication and the Community." Scientific American 227 (September 1972): 142-51. Hanneman, Gerhard J. "Communication Substitutes for Transportation: The Movement of Information Versus the Movement of PeOple." Los Angeles: University of Southern California, 1974. (Mimeographed.) . "A Computer Simulation of Information Dif- fusion in a Peasant Community." Master's thesis, Michigan State University, 1969. Hartman, John J., and Walsh, James A. "Simulation of Newspaper Readership: An EXploratory Analysis of Social Data." Social Science Quarterly 49 (1969): 840-52. Heirich, Max. "The Use of Time in the Study of Social Change." American Sociological Review 29 (June 1964i: 386-97. Heiss, Klaus P.; Knorr, Klaus; and Mergenstern, Oskar. Long Term Projections of Power: Political, Economic, and MilitaryForecasting. CamBridge: Ballinger Publishing Company, 1973. Hemmens, George C. "Analysis and Simulation of Urban Activity Patterns." Socio-Economic Planning Sciences 4 (1970): 53-66. Hinshaw, Mark L. "Wiring Megalopolis: Two Scenarios." In Communications Technology and Social Policy, pp. 305-18. Edited by G. Gerbner, L. Gross, and W. Melody. New York: John Wiley & Sons, 1973. 293 Holman, Mary A. "A National Time-Budget for the Year 2000." Sociology and Social Research 46 (October 1961): 17-25. Hughes, Ann, and Grawoig, Dennis. Statistics: A Foun- dation for Analysis. Reading, Mass.: AddiSon- Wesley, 1971. Hyman, Herbert H. Secondary Analysis of Sample Survgys: Principles, Procedures and Potentialities. New York: John Wiley & Sons, 1972. Jacoby, Joan E., and Harrison, Stephen. "Multi-Variable Experimentation and Simulation Models." In Computer Simulation of Human Behovior, pp. 626-’ 37. Editediby J. Dutton and W. Starbuck. New York: John Wiley & Sons, 1971. Johnson, F. Craig, and Klare, George R. "General Models of Communication Research: A Survey of the Developments of a Decade." Journal of Communi- cations 11 (March 1961): 13-45. Johnson, M. Bruce. "Travel Time and the Price of Leisure." Wgstern Economic Journal (Spring 1966): 135-45. Jolson, Marvin A., and Rossow, Gerald L. "The Delphi Process in Marketing Decision Making." Journal of Marketing Research 8 (November 1971): 443-48. Kahn, Ephraim. "Commercial Uses of Broadband Communi- cations." In Talking Back: Citizen Feedback and Cable Technology, pp. 247-65. Edited by I. Pool. Cambridge: The M.I.T. Press, 1973. Katzman, Natan. "Communication Flow and Social Entropy." Ekistics 29 (March 1970): 212-19. . "The Impact of Communications Technology: Promises and Prospects." Journal of Communi- cation 24 (Autumn 1974): 47-58. Kline, F. Gerald. "Theory in Mass Communication Research." In Current Perspeotives in Mass Communication_Research, Vol. I., pp. 17-40. Editediby F. Kiine aid P. Tichenor. Beverly Hills: Sage, 1972. Kotler, Philip, and Schultz, Randall L. "Marketing Sys- tem Simulations." In Simulation in Social and Administrative Science, pp. 481-572. Edited by H. Guetzkow, P. Kotler, and R. Schultz. Englewood Cliffs, N.J.: Prentice-Hall, Inc., 1972. 294 Kramer, John F. "A Computer Simulation of Audience Exposure in a Mass Media System: The United Nations Information Campaign in Cincinnati 1947-48." Ph.D. dissertation, M.I.T., 1969. Kurtz, David L.; Boone, Louis E.; Johnson, James C.; and Bonno, John A. "City Shoppers and Urban Identif- ication Revisited." Journal of Marketing 38 (July 1974): 67-68. Lancaster, Kelvin. "Allocation and Distribution Theory: Technological Innovation and Progress." American Economic Review 56 (May 1966): 14-23. Linder, Staffan Burenstam. The Harried Leisure Class. New York: Columbia University Press, 1970. Little, John D. C., and Lodish, Leonard M. "A Media Planning Calculus." Operations Research 17 (January-February 1969): 1-33. Mabry, Bevars D. "An Analysis of Work and Other Con- straints on Choices of Activities." Western Economic Journal 8 (September 1970): 213-25. Mandelbaum, Seymour J. Community and Communication. New York: W. W. Norton & Company, 1972. Martino, Joseph P. Technological Forecasting for Decisionmaking. New York: American Elsevier Publishing Company, Inc., 1972. ' Mauser, Ferdinand F. "A Universe-In-Motion Approach to Marketing." In Managerial Marketing: Perspec— tives and VieWpoints, pp. 48436. Edited by E. Kelley and W. Lazer. Homewood: Richard D. Irwin, 1967. McNemar, Quinn. PsychologicalStatistics: Third Edition. New Yofk: John Wiley & Sons, I962. McVoy, D. Stevens, and Reynolds, Richard. "Cost Barrier Cracked in Two-Way Cable TV." Electronics (February 20, 1975): 96-100. Meadows, Donella H.; Meadows, Dennis L.; Rauders, Jorgen; and Behrens, William W., III. The Limits to Growth. New York: Universe Books, 1972. Meier, Richard L. A Communication Theoryof Urban Growth. Cambridge: Tfie M.I.T. Press, 1962. 295 Meier, Richard L. "Human Time Allocation: A Basis for Social Accounts." Journal of the American Institute of Planners 25 (1959): 27. . "Urban Ecostructures in a Cybernetic Age: Responses to Communication Stress." In Communi- cations Technology and Social Poligy, pp. 353-64. Edited by G. Gerbner, L. Gross, and W. Melody. New York: John Wiley & Sons, 1973. Milgram, Stanley. "The Experience of Living in Cities." Ekistics 30 (August 1970): 145-50. Moore, Wilbert E. Man, Time, and Society. New York: John Wiley & Sons, T963. Naylor, Thomas H.; Balintfy, Joseph L.; Burdick, Donald S.; and Chu, Kong. Computer Simulation Technigues. New York: John Wiley & Sons, 1966. Naylor, Thomas H. "Methodological Considerations in Simulating Social and Administrative Systems." In Simulation in Social and Administrative Science, pp. 647-72. Edited by H. Guetzkow,. P. Kotler, and R. Schultz. Englewood Cliffs, N.J.: Prentice-Hall, 1972. Newell, William T., and Meier, Robert C. "Business Systems Simulation." In Simulation in Social and Administrative Science, pp.i4IS-58. Edited by H. Guetikow, P. Kotler and R. Schultz. Englewood Cliffs, N.J.: Prentice-Hall, 1972. Nuckols, Robert C., and Mayer, Charles S. "Can Indepen- dent Responses Be Obtained from Various Members in a Mail Panel Household?" Journal of Marketing Research 7 (February 1970): 90-94. Ohls, James C. "Marginal Cost Pricing, Investment Theory and CATV." The Journal of Law and Economics 13 (1970): 439-60. . "Marginal Cost Pricing, Investment Theory and CATV: A Reply." The Journal of Law and Economics 14 (October 19717: 517-19. On the Cable: The Television of Abundance. Report of the Sloan Commission on Cable Communication. New York: McGraw-Hill, 1971. 296 Oort, C. J. "The Evaluation of Travelling Time." Journal of Transport Economics and Policy T§eptember 1969): 279-86. Park, Rolla Edward. "The Growth of Cable TV and Its Parker, Peters, Probable Impact on Over-The-Air Broadcasting." American Economic Review 61 (May 1971): 69-73. . Potential Impact of Cable Growth on Tele- vision Broadcasting. Santa Monica, Calif.: Rand Corporation, 1970. Edwin B. "Technology Assessment or Institu- tional Change?" In Communications Technology and Social Policy, pp. 533-48} Edited by G. Geaner, L. Gross, and W. Melody. New York: John Wiley & Sons, 1973. William H., and Ford, Neil M. "A Profile of Urban In-Home Shoppers: The Other Half." Journal of Marketing 36 (January 1972): 62-63. Pool, Ithiel de Sola. "Computer Simulation of Total Societies." In The Study of Total Societies, pp. 45-65. Edited by S. Klausner. New York: Frederick A. Praeger, 1967. . "The Rise of Communications Policy Research." Journal of Communications 24 (Spring 1974): 31-42 a ; Abelson, Robert P.; and Popkin, Samuel L. Candidates, Issues and Strategies: A Computer Simulation of the 1960 and 1964 Presidential Elections. Cambridge: M.I.T. Press, 1965. ; Murray, Charles; and Dobb, Kenneth. "Prospects for On-Demand Media." In Talking Back: Citizen Feedback and Cable Technology, pp. 266-316. Edited by I. Pool. Cambridge: The M.I.T. Press, 1973. Popkin, Samuel L. "A Model of a Communication System." American Behavioral Scientist 8 (1965): 8-11. Ports, Delmer C. "Trends in Cable TV." IEEE Trans- actions on Communications COM-23 (January 1975): 92-96. Price, Monroe E. "The Illusions of Cable Television." Journal of Communication 24 (Summer 1974): 71-76. 297 Pritsker, A. Alan B., and Kiviat, Philip J. Simulation with GASP II. Englewood Cliffs, N.J.: Prentice H511, I969. Rathmell, John M. "Discretionary Time and Discretionary Mobility." In Managerial Marketing: Pergpectives and Viewpoints, pp. 145-55. Edited by E. Kelley and W. Lazer. Homewood: Richard D. Irwin, 1967. Ray, Michael L. "A Decision Sequence Analysis of Develop— ments in Marketing Communication." Journal of Marketing 37 (January 1973): 29-38. Reid, Alex. New Directions in Telecommunications Research. A Report Prepared for the Sloan Com- mission on Cable Communications, 1971. Reynolds, Fred D. "An Analysis of Catalog Buying Behavior." Journal of Marketing 38 (July 1974): 47.51. F Rivkin, Steven R. Cable Television: A Guide to Federal Regulations. Santa Monica, Calif.: Rand Cor- poration, 1973. Robinson, John P. "Television and Leisure Time: Yester- day, Today, and (Maybe) Tomorrow." Public gpinion_guarterly 33 (Summer 1969): 210-22. , and Converse, Philip E. "The Impact of Tele- vision on Mass Media Usages: A Cross-National Comparison." In The Use of Time, pp. 197-212. Edited by A. Szalai. The Hague: Mouton, 1972. Rose, Kenneth, and Stevens, Richard K. "Design of a Switched Broad-Band Communications Network for Interactive Services." IEEE Transactions on Communications COM-23 (January 1975): 49-55. Sardella, Vincent. "Interactive Cable: Federal Policy and Programs." IEEE Transactions on Communi- cations COM-23 (January 1975): 171-77. Schary, Philip B. "Consumption and the Problem of Time." Journal of Marketing 35 (April 1971): 50-55. Schneider, Annerose. "Patterns of Social Interaction." In The Use of Time: pp. 317-34. Edited by A. Szalai. The Hague: Mouton, 1972. 298 Schultz, Don E.; Block, Martin; and Jacobowitz, Bret. DONMAR I: A Computor Simulated Media Buying Game. Department of Advertising, Michigan State University, East Lansing, Michigan, 1975. Schultz, Randall L., and Sullivan, Edward M. "Develop- ments in Simulation in Social and Administrative Science." In Simulation in Social and Adminis- trative Science, pp. 3-50. Edited by H. Guetzkow, P. Kotler, and R. Schultz. Englewood Cliffs, N.J.: Prentice-Hall, 1972. Smith, Ernest K. "Pilot Two-Way CATV Systems." IEEE Transactions on Communications COM-23 (January l975): 111-20. Smith, Ralph Lee. The Wired Nation. New York: Harper & Row, 1972. Sorokin, Pitirim A., and Berger, Clarence Q. Time- Budgets of Human Behavior. Cambridge: Harvard University Press, 1939. Speh, Thomas W. "The Performance of a Physical Dis- tribution Channel System Under Various Conditions of Demand Uncertainty: A Simulation Experiment." Ph.D. dissertation, Michigan State University, 1974. Staikov, Zahari. "Time-Budgets and Technological Pro- gress." In The Use of Time, pp. 461-82. Edited by A. Szalai. The Hague: Mouton, 1972. Stidsen, Bent, and Schutte, Thomas F. "Marketing as a Communication System: The Marketing Concept Revisited." Journal of Marketing 36 (October 1972): 22-27. Stigler, George J. "The Economics of Information." The Journal of Political Economy 69 (June 1961): 213-25. “c Stone, Philip J. "Computer Models of Information Pro- cessing." In Personality Theory and Information Processing. Edited by H. Schroder and P. Suedfeld. New York: The Ronald Press, 1970. . "Models of Everyday Time Allocations." (In The Use of Time, pp. 179-90. Edited by A. Szalai. The Hague: Mouton, 1972. 299 Stone, Philip J. "The Analysis of Time-Budget Data." In The Use of Time, pp. 89-112. Edited by A. Szalai. The Hague: Mouton, 1972. Summary of United States Time Use Survey. Institute for SoCial Research, University of Michigan, Ann Arbor, Michigan, 1966. Szalai, Alexander. "Introduction: Concepts and Practices of Time-Budget Research." In The Use of Time, pp. l-12. Edited by A. Szalai. The Hague: Mouton, 1972. . "The lZ-Nation Comparative Study." In Technology, Human Values and Leisure. PP. 164- 90. Edited by M. Kaplan and Philip Bosserman. Nashville: Abingdon Press, 1971. , ed. The Use of Time. The Hague: Mouton, 1972. . "Trends in Comparative Time-Budget Research." The American Behavioral Scientist (December 1966): 3-8. Tauber, Edward M. "Why Do People ShOp?" Journal of Marketing 36 (October 1972): 46-49. Thorelli, Hans B. "Concentration of Information Power Among Consumers." Journal of Marketing Research 8 (November 1971): 427-32. Tomlinson, Janet; Bullock, N.; Dickens, P.; Steadman, P.; and Taylor, E. "A Model of Students' Daily Activity Patterns." Environment and Planning 5 (1973): 231-65. Tsokos, Chris P. Probability Distributions: An Intro- duction to Prohébility Theory_with Applications. Belmont, Calif.: Wadsworth PubliShing Company, 1972. User's Guide: The 6500 Scope/Hustlergperating System. Vol. II. Computer Laboratory, Michigan State University, East Lansing, Michigan. Wagenheim, George D. "The Performance of a Physical Distribution Channel System Under Various Con- ditions of Lead Time Uncertainty: A Simulation Experiment." Ph.D. dissertation, Michigan State University, 1974. 300 Ward, John B. "What Belongs on the Cable?" In Talking Back: Citizen Feedback and Cable Technology, pp. 139-90. Edited by I. Pool. Cambridge: The M.I.T. Press, 1973. Young, Kan-Hua. "The Demand and Supply of Telecommuni- cation Services in the United States: Empirical Results." In Long Term Projections of Power, pp. 193-206. Edited by K. Heiss, K. Knorr, and O. Morgenstern. Cambridge: Ballinger Publishing, 1973. Zehna, Peter W. Probability Distributions and Statistics. Boston: Allyn and Bacon, Inc., 1970.