National Academies Press: OpenBook

Diffusion Processes and Fertility Transition: Selected Perspectives (2001)

Chapter: 3 Diffusion in Sociological Analysis

« Previous: 2 Potatoes and Pills: An Overview of Innovation-Diffusion Contributions to Explanations of Fertility Decline
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

3
Diffusion in Sociological Analysis

ALBERTO PALLONI

OBJECTIVES

There are a number of very lucid, thorough, and authoritative reviews about the nature and applications of theories and models of diffusion in sociology (see, for example, Rogers, 1962, 1973, 1988, 1995; Valente, 1995). However, this literature is neither geared to deal with problems in the explanation of demographic phenomena nor does it indicate how to take advantage of new developments in economic and social network theories and methodological innovations for the study of dynamic processes. This paper is designed to fill this gap. In particular, I have four interrelated goals:

  1. To identify the backbone of diffusion models and theories in sociology, and to show that recent formulations and applications require robust, well-specified theories about social systems and about the positions that individuals exposed to diffusion occupy within the social structure;

  2. To illustrate recent applications of diffusion models and theories in two key areas of sociology, social movements and social organizations;

  3. To define conditions (“identification conditions”) for testing new hypotheses and conjectures that invoke diffusion processes. These conditions are strict, are difficult to satisfy, and have implications for issues ranging from data collection to selection of estimation procedures. I argue that unless these conditions are met, we will not be able to identify

Alberto Palloni is professor of sociology at the University of Wisconsin, Madison.

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

diffusion processes from among other processes producing similar observable outcomes.

  1. To argue that until very recently at least, applications of diffusion models in demography have not taken advantage of innovations identified in goal 1, and have not adhered to the formal conditions identified in goal 3. Thus, these applications are unlikely to be of much help to improve our understanding of demographic phenomena.

The organization of the paper is as follows: the first sections deal with goals 1 and 2, respectively, middle sections focus on goal 3, the next section discusses material related to goal 4 and, the last section contains a summary and concluding remarks.

THE BASIC MODEL OF DIFFUSION IN SOCIOLOGY

In this section I show that sociological theories of diffusion have evolved from fairly simple propositions regarding average or aggregate behavior into complex formulations about how individuals define preferences and make decisions to realize those preferences. In this section I argue that in order to be analytically useful, diffusion models require theorizing about social structures, about the positions that individuals occupy in them, about individual decision-making processes that accompany adoption of a behavior, and about the constraints these individuals face. I conclude that it is unilluminating to confront diffusion theories with competing explanations that regard behaviors as responsive to “structural” factors, such as socioeconomic positions or social class membership, as if diffusion processes did not require or could proceed independently of structural factors that characterize the environment where individuals act and where behaviors take place. Similarly, it is misleading to cast diffusion models or theories against alternative ones on the grounds that the latter are usually erected on a foundation of assumptions about rational actors and well-defined decision-making processes, as if diffusion processes did not require making assumptions about preferences, costs, and a rational calculus. Well-defined diffusion hypotheses and models must be built on assumptions about social and economic conditions that constrain individual actors’ preferences and resources, and rely on these assumptions no less than alternative hypotheses and models often pitted against them. This is not to say that diffusion models or theories do not have a specificity of their own. They do, and it will be the task of the next sections to identify what this specificity is. In the end, however, my message is somewhat pessimistic because the conditions for identification of a diffusion process from observables are fairly hard to meet, much harder than what is normally implied in traditional applications of diffusion theories and models to sociological and demographic analyses.

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

Diffusion Explanations and Structural Explanations

“Structural” explanations of behavioral changes seek their cause in the alteration of preferences and opportunities that result from either changes in positions that individuals occupy (individual social mobility) or from reshuffling of resources associated with a given social position (structural social mobility or redistribution of wealth). Diffusion explanations or models, on the other hand, attempt to identify a cascading mechanism that leads to cumulative adoption of behaviors by some individuals, even while their social position, or the resources associated with them, changes only trivially or remains unaltered. In diffusion models, the behavior “spreads” and is adopted by individuals irrespective of their socioeconomic positions, even among those whose social or economic positions are hypothetically associated with cost-benefit calculations that do not necessarily require the new behavior. Adopting the new behavior occurs as a result of reevaluation of one’s own choices in light of other people’s behavior, not as a strategic response or accommodation to a realignment of resources associated with one’s social position in the social system. To use the terminology Coleman (1990) coined for the study of collective behavior, diffusion models are built on the central idea that individuals transfer partial or total control of their own behavior to others. As I will show later, this requires a decision process as complicated (or uncomplicated) as the ones that are normally associated with structural explanations.

Diffusion processes do not always involve adoption of new behaviors. In fact, they may include abandonment of a recently adopted behavior or resistance to change. For example, it has been observed that, contrary to expectations, class-based political alignments do not always take hold at a pace that is commensurate with advances of industrialization. Instead, traditional political allegiances, based on language or ethnic identities, may remain dominant long after industrialization has created the structural conditions for class-based politics. This type of phenomenon has been studied widely in political sociology to understand the stubborn persistence of nonclass-based allegiances and ethnic enclaves (Hechter, 1975). In these cases observed individual political behavior (voting behavior) is at odds with what is expected by virtue of an individual’s position or ranking in the social system. Failure of individuals to act according to class positions—an expectation derived from a “structuralist” explanation of political behavior—occurs as a result of adherence to practices that were consistent with positions occupied prior to the social and economic transformations that accompanied industrialization. What is diffused or adopted here is the individual resistance to act according to class-based principles (the new behavior), and the reinforcement of tradi-

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

tional political alignments (the old behavior). If political sociologists were able to gather information on collective protests against British rule, rather than just on voting patterns, they would observe waves of protests extending across and confined within the boundaries of the British fringe, much as they observe waves of protests in the United States during the 1960s (Myers, 1997).

Similarly, we know all too well that fertility decline in Europe did not always follow a trajectory consistent with social and economic transformations that accompanied industrialization. Instead, the course of the decline revealed a marked tendency to proceed along or be halted by ethnic, language, and religious boundaries. The resulting geographic and territorial clustering of fertility levels and patterns has been construed as evidence against a structural explanation of fertility decline, and as support for the hypothesis that fertility changes were strongly associated with ideational or cultural changes and diffusion mechanisms.1 The existence of strong clustering of fertility levels along cultural lines could be evidence of either diffusion of a new behavior (adoption of contraception and a low fertility norm) in areas with lower than expected fertility (structural changes), or of resistance to the new behavior (rejection of birth control and adherence to a high fertility norm) in areas with higher than expected fertility.

The foregoing examples share two features. The first is that in both cases we establish a contrast between an explanation that infers an expected behavior from a reading of individual socioeconomic positions (the structuralist explanation) with an alternative explanation that infers a pattern of expected behavior from the likely adherence of actors to ethnic, religious, or cultural prescriptions or beliefs shared by others in the same community, including individuals belonging to different social classes or occupying different socioeconomic positions. In the latter case, the likelihood of adherence to prescriptions increases as a function of others’ adherence to it (or others’ resistance to the novel behavior). The definition of what is included in “others” is and must be a key element of the theory, as should the identification of the mechanisms that reproduce efficiently adherence to prescriptions and beliefs.

The second common feature shared by these two examples is that the structuralist or socioeconomic explanation and the diffusion explanation offered to account for the phenomena rest on the idea that individuals are decision makers, acting in uncertain environments; sorting through limited information on prices, utilities, constraints, and potential outcomes of alternative behaviors; elucidating their own preferences; and ultimately taking some course of action. But, whereas investigators are normally careful to produce a thorough definition of the decision process associated with the structuralist explanation, they all too often fail to specify the

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

decision-making process associated with diffusion, to the point that this appears, in many instances, as a result of passive contagion and the irrational or at least a-rational adoption of a behavior. This is a situation not unlike the one found until recently in the study of collective actions that could be explained only through recourse to the irrationality of actors (Coleman, 1990). The exceptions to this lack of attention to decision-making processes embedded within diffusion are precisely the most recent studies and formulations of diffusion processes in sociology, economics, and demography (Montgomery and Chung, 1994; Montgomery and Casterline, 1993; Valente, 1995; Marsden and Friedkin, 1993; Burt, 1987).

Lack of theoretical specificity is not the only problem we face as we try to identify diffusion processes. In fact, most of the evidence produced in sociology and demography to distinguish between explanations based on diffusion arguments from those attributing the primary role to socioeconomic or structural changes is carved out of aggregate, not individual, data. Because the individual adoption process is never defined, the aggregate process is also ill conditioned: there is rarely a way to determine what kind of aggregate evidence one would expect when the individual adoption process is left unspecified. This leads to the very generalized practice of using residual evidence or, equivalently, to infer the validity of a diffusionist explanation from the failure of the structural explanation: the explanatory power assigned to the diffusion argument is always directly proportional to the magnitude of the inconsistency between observed outcomes and those expected from a competing structural explanation. Handling only aggregate and residual evidence leads to the central problem in this literature—both in sociology or demography— namely, the inability to identify the key process from observables.

The Elements of an Explanation Based on Diffusion Processes

A classic definition of diffusion is the following: “(Diffusion) is the process by which an innovation is communicated through certain channels over time among the members of a social system. It is a special type of communication, in that the messages are concerned with new ideas” (Rogers, 1983:19). There are a number of essential elements contained in this definition: the innovation, the population of potential adopters, those who adopt, and the mechanisms through which adopters and potential adopters communicate with each other. The classical problem in diffusion models is to understand who adopts the innovation, and how fast they do so. Thus, Rogers (1995) distinguishes different types of adopters depending on how early during the adoption process occurs. To these groups one could add a category including those who never adopt, much as in social mobility we recognize movers and stayers. Delays in adoption or resis-

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

tance to adopt are explained by inadequate information or by uncertainty about the results or outcomes associated with the innovation. As the process advances and more individuals adopt, and as the outcomes of adoption by others become observable, more individuals’ resistance to adoption crumbles as the information is enriched and their uncertainty about risks, costs, and benefits diminish.

Although later in the paper I will introduce a more complex notion of diffusion, in the remainder of this section I will focus on the classic definition just given. I will use it as a reference to identify elements of a diffusion process that should be important in model building but that many applications overlook. The simplicity of the classic definition is deceiving, for it contains explicitly or implicitly a number of key elements that are important to identify at the outset. First, diffusion occurs through an individual decision-making process in which there are costs and benefits (and implicitly preferences) associated with adoption (or its obverse, resistance to adoption), as well as information and ignorance about prices, costs, outcomes, and alternatives. In their influential work on cultural transmission, Cavalli-Sforza and Feldman (1981) stress the importance of decision making as the factor that distinguishes cultural from biological evolution. Whereas the latter is driven by natural selection (or genetic drift), the former is characterized by the influence of individual decision making that may reinforce or offset the pressures of natural selection: “In cultural evolution, however, there is in addition [to natural selection] a second mode of selection, which is the result of the capacity of decision making” (Cavalli-Sforza and Feldman, 1981:10).

Diffusion only occurs because individuals decide to adopt after observing others do so, and after updating their information by including observed outcomes associated with others’ adoption into their own decision-making process. There may be a variable number of stages in this decision-making process (Rogers, 1983), but what is important is that its core is an individual who is making cost-benefit calculations under uncertainty about whether to join others in adopting a behavior or, alternatively, resisting. A diffusion model rests on assumptions and imageries not dissimilar to the ones that prevail when, for example, we refer to individuals changing their fertility behavior as a result of socioeconomic changes that affect them (the so-called demand theories of fertility). The vast majority of applications of diffusion models in both demography and sociology neglect this very simple tenet of diffusion models: adopters and nonadopters are rational decision makers and adoption is the outcome of a rational decision-making process. These issues have been confronted head-on in only a handful of applications. For example, in a recent study Montgomery and Casterline (1996) define three distinctive elements of a diffusion process—social learning, social influence, and institutional con-

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

straints—which operate to determine and shape individual decision making about adoption of behaviors. Similarly, Erbring and Young (1979) and Marsden and Friedkin (1993) carefully elaborate on the types of social relations that are relevant for processes whereby behaviors of one individual are affected by consideration of behaviors of other individuals belonging to the same group or social system. Coleman’s (1990) study of collective action and those involving or generating trust reveal the fundamental elements of the decision-making process on which every diffusion process depends. Even in the study of organizations and organizational diffusion (DiMaggio and Powell, 1991), there is explicit consideration of actors who imitate organizational features adopted by successful organizations as a device to minimize uncertainty.

Second, given conditions defining their preferences and opportunities, individual decision makers may be more or less resistant both to adopt innovations and, if they adopt, more or less reluctant to jettison the innovation from the menu of practices and behaviors they normally employ. That is, after one accounts for all elements entering in the decision to adopt or to resist, there might be individuals who are more (less) risk averse and adopt more (less) easily than others. These will be forerunners (laggards) in the diffusion process (Rogers, 1983). As stated by Cavalli-Sforza and Feldman (1981:39), “It seems very likely, a priori, that there is variation between individuals in their capacity both to learn of an innovation and to decide for adoption. Many factors contribute to such variation, including social and economic stratification, geographic conditions such as means of transportation, availability of communication networks, and, last but not least, individual differences in the behavioral characteristics that govern both awareness and eventual adoption.” This acknowledges that after accounting for a number of social and economic factors, we are likely to face the existence of “unmeasured heterogeneity” or the inability to include all elements that contribute to the individual’s decision regarding the innovation. It is a concept analogous to frailty in the analysis of mortality and induces the same empirical patterns: as individuals who are more resistant to adopting become a larger fraction of the pool of nonadopters, the overall risk of adoption will tend to decrease. But this is not a reflection of a risk profile of adoption that decreases over time. Rather, it is an artifact of the changing composition of the pool of nonadopters as the process progresses over time. To my knowledge, the traditional literature on diffusion processes in sociology or demography has not addressed the problem created by the unmeasured resistance to adoption, except insofar as the study of forerunners and the conditions that determine their appearance is indeed a way to identify factors influencing unmeasured resistance.2 In general, however, we neglect the issue altogether. This practice is explained by one of two factors: either the assump-

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

tion is made that all relevant factors were well measured (including those affecting awareness and propensity to adopt), or the focus of attention is on aggregate patterns of adoption. It is only recently, mainly through the influential work of Granovetter (1978) and Valente (1995), that the concept of individual (or group) thresholds has been introduced as a way to handle the problem, but still without deriving the full consequences for model testing. Later I will provide an interpretation, by no means unique, of unmeasured resistance to adoption.

Equally important for the successful progression of diffusion are processes that may undermine continued practice of the new behavior. To the extent that these acquire some dominance, individuals are more likely to abandon the new practice or behavior some time after adoption. Despite the fact that this is a rather key part of a diffusion process, it is rarely mentioned and almost never explicitly modeled or studied.3

Third, the decision-making process underlying adoption of new behaviors occurs within a social structure composed of formal and informal elements. Individuals occupy positions within these social structures, perform certain roles, and are connected formally and informally to a number of other individuals within them through relations of authority, functional rapports, respect, and trust. They adhere to values and norms that shape preferences, constrain the field of feasible behaviors, and alter the information they may receive about prices, utilities, and ultimately about what others are doing. Despite the fact that often it is difficult to tell so from actual empirical research involving diffusion models, diffusion processes are affected by the social structure of systems within which they are occurring. Social structures determine the content and shape of the repertoire of feasible behaviors (“Is the behavior within the realm of conscious choice?”), individual’s preferences (“Is the behavior advantageous at all?”), and individual’s resources (“Can individuals adopt at low costs?”). The questions within quotes describe Coale’s well-known desiderata for fertility change (Coale, 1973; see also Lesthaeghe and Vanderhoeft, this volume) and could be utilized equally well by an explanation resting on diffusion as an alternative mechanism involving adjustment to structural changes. I will elaborate on this in later sections.

The importance of social structure appears to weigh more heavily when the diffusion process is suspected to be under the control of internal sources rather than external sources of diffusion. However, even the idea that external sources of diffusion have an impact independent of individuals’ position in the social structure is acceptable only as a tool to render the algebra of models tractable, but it is woefully inadequate for analytic purposes. Some of the best original work on diffusion processes emphasizes that social diffusion is an analytically sterile construct if not cast against a social structure: “It is as unthinkable to study diffusion

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

without some knowledge of the social structure in which potential adopters are located as it is to study blood circulation without adequate knowledge of the structure of veins and arteries” (Katz, 1961; cited in Rogers, 1983:25). Similarly, in their influential study on use of hybrid corn among farmers in two Iowa communities, Ryan and Gross (1943) argue that it is the social structure that may explain the delay with which certain technologies are adopted. They reason that, if all individuals act as rational actors, adoption of an advantageous technological innovation must occur instantaneously and simultaneously. Delays and lags in the process and the emergence of laggards in the population of potential adopters can only be explained by institutional constraints and by sociocultural and psychological factors that influence the diffusion process. In this case, social structure is taken to be an obstacle rather than a facilitator. Structure accounts for the slow progress of diffusion rather than diffusion undermining the constraints fabricated by social structures.

Although emphasis on the importance of social structure for diffusion processes is hardly new, and even despite the fact that there are good examples demonstrating careful attention to social structure (Rogers and Kincaid, 1981; Coleman et al., 1966; Burt, 1987), it has seldom been systematically incorporated into actual empirical research. It is only recently that sociologists interested in diffusion have begun to pay close attention to it and accounting for it explicitly in the formulation of models. In a recent paper, Strang and Soule (1997:1) make the point that while diffusion studies inquire about how practices spread, they also “provide an opportunity to locate and document the social structure, where we consider how patterns of apparent influence reflect durable social relations.” Furthermore, because these models involve individual decision making subjected to constraints imposed by a social structure, they may”… verge on the one hand towards models of individual choice, since diffusion models often treat the adopter as a reflective decision maker… [or] verge on the other [hand] towards a broader class of contextual and environmental processes, where conditions outside the actor shape behavior” (Strang and Soule, 1997:2).

Fourth, once innovations are adopted, they could be abandoned and replaced by other technology, instruments, or behaviors. Thus, in addition to understanding who adopts and how fast they do so, models of diffusion should specify the obverse process, the persistent use of the innovation. This aspect of a diffusion process is of importance in applications to social behaviors that are inherently reversible or unstable. For example, participation in mass protests usually involves increased risk of participation followed by increased risk of withdrawal from the pool of protesters. Withdrawal from protest is as much a diffusion process as is participation in it (Myers, 1997), and could be triggered and encouraged

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

by external reprisals. Discontinuation is also relevant for situations where what is at stake is the adoption of an innovation such as contraception. Contraceptive discontinuation is an obvious illustration that has become a staple of empirical studies of contraception, but so is the possibility that certain groups may adopt contraception and then abandon altogether the very ideal of family limitation. If one succeeded in providing a convincing explanation of fertility decline in Western Europe entirely based on diffusion arguments, we should also explain why the decline turned out to be irreversible. Although this seems an obvious requirement, I have seen no systematic evidence indicating that the issue has been raised, much less treated systematically (for an exception, see Kohler, 1997). Note that this is not a requirement that applies to explanations invoking adaptation to new social and economic conditions. Whenever possible and nontrivial, an ideal diffusion model ought to specify the conditions for the persistence of adoption.

Fifth, the social and economic environment may be modified by the process of adoption itself, and may involve feedbacks accelerating or retarding the process. The adoption of some computer technologies, for example, becomes unavoidable once a critical mass of users has adopted because the incentive structure for all users is altered, becomes more favorable for adoption of the technology, and creates niches for the introduction of even newer technology. The adoption of operating systems for PCs proceeds in this fashion, with software production being the element that induces interdependence among consumers in the market. Similarly, changing prices of a product induced by partial adoption of a technological innovation in agriculture will alter the elements that enter into the calculus of nonadopters (Ryan and Gross, 1943; Hagerstrand, 1967). Adoption of organizational features such as civil service reform may begin to occur for reasons that have more to do with the establishment of legitimacy of the practice than with associated increases in efficiency (DiMaggio and Powell, 1991). Adoption of a practice may accelerate as organizations that have not yet adopted find it advantageous to mimic what others have done successfully as a way to sharpen their competitive edge in the new environment created by a handful of forerunners (Fligstein, 1985). DiMaggio and Powell’s “mimic processes,” whereby organizations imitate what other organizations do, refers to processes whereby the linkage between a practice and its net benefits is subject to less variability, but also to processes where the institutional environment is so changed by early adopters that adoption simply becomes more cost effective. Only the latter is an example of endogenous feedback.

Similar processes may be at work in fertility behavior: forerunners who first adopt fertility control not only generate an environment with reduced uncertainty for others to follow, but may also create emulation

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

conditions. This can happen, for example, if with fewer children they are able to support higher or better educational standards and if, in the long run, this enhances their power and prestige. To the extent that this is so, nonadopters pursuing power and prestige will be better off if they imitate fertility limitation. As the process evolves, the institutional context to satisfy the demand for more and better education also evolves, thus changing the context in which fertility decision making is taking place.

In the case of organizational adoption, the pool of means to attain some ends is changed by adoption of newer procedures or strategies, and so is the ranking of those that are preferred among all organizations in the field, not just those who initially adopt. In the case of fertility, the connection between fertility limitation and power and prestige via children’s education converts the adoption of contraceptive behavior from an oddity to a useful and productive behavioral strategy.

In these examples taken from sociology of organizations and fertility, there is endogenous feedback since the spread of the behavior changes the elements that enter into the decision-making process of everybody, including nonadopters. Surely, there must be considerable empirical variability in the lags with which the feedback operates, and in their actual significance for individual decision makers. Thus, endogenous feedback need not be an inherent nor a uniform characteristic of all diffusion processes. But, when it is, it will alter individual probabilities of adoption for individuals who have not yet adopted at a certain time in the process.4

The combination of some of these five elements of a diffusion process may produce lightning-fast spread of innovations. By the same token, though, particular constellations of the elements may lead to excruciatingly slow adoption, to innovation processes that begin rapidly but then taper off without ever reaching near saturation, or to those that fail altogether and are then relegated to the pool of diffusion processes that we will never be able to study.5 An immediate corollary of this inherent variability is that it is not necessarily correct to infer the existence of a causal mechanism (diffusion mechanisms versus structural mechanisms) only from observation of the relative speed with which a behavioral change occurs. It is as much an error to believe that when a process of behavioral change is quick and swift it must have been due to diffusion as it is to think that no diffusion process could be responsible for slow changes. The observed rate of change in the prevalence of a behavior by itself will generally be of limited help to identify a diffusion process because the effects of the basic elements of a diffusion process may lead to outcomes that can also be produced by mechanisms not associated with diffusion at all. Rapid rates of change in a behavior in the absence of changes in structural condition may be a reflection of diffusion, but it surely should not be taken as prima facie evidence of its existence or predominance.

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

These five elements are strategic for a proper formulation of a diffusion model. But, needless to say, they are not always taken into account. In one subsequent section, I show that this oversight leads to shortcomings in sociological applications. In another section, I revisit these elements and use them to define more formally the nature of a diffusion process and the mechanisms through which it operates.

DEVELOPMENT OF DIFFUSION MODELS IN SOCIOLOGY

In this section I discuss developments in the formulation and application of diffusion models in sociology. I start with early models that mimic those used for the study of the spread of diseases and focus on narrow aggregate outcomes. I then discuss some of the most novel applications in the areas of collective action and organizations.

Early Studies and Formulations

The main territory of diffusion theories and models is the innovation. Innovations by their very nature require communication and information for adoption. They are also risky because their outcomes are for the most part uncertain or unknown, thus requiring an agent engaged in a decision-making process. It is not surprising, then, that diffusion processes have been mostly used to study adoption of innovations. The most influential works include Ryan and Gross’s (1943) analysis of the diffusion of hybrid corn, Hagerstrand’s (1967) investigation of the diffusion of tuberculosis tests in Sweden, Coleman et al’s (1966) study of the adoption of tetracycline among Midwestern doctors, Katz and Lazarsfeld’s (1955) celebrated formulation of the two-step flow of influence process, and Rogers’ and Kincaid’s (1981) analysis of contraceptive behavior. The main goal of all these studies is to assess the effects of the mass media, the degree of influence of individuals located at the top of the community hierarchy (agents of change), and the relative contribution of interpersonal interactions within the boundaries of the community where the innovation is spread. In all these applications, the empirical evidence gathered to demonstrate the existence of diffusion includes individuals, and their characteristics and interactions. In only one study was the evidence restricted to observation of aggregate outcomes such as proportion of adopters. With the exception of the works by Hagerstrand and Ryan and Gross, these studies placed emphasis on interpersonal relations and channels of influences as the mechanisms fostering or impeding diffusion. In this sense they anticipated some of the most useful work on social influences in general and diffusion in particular (Bandura, 1986; Moscovici, 1985; Marsden and Friedkin, 1993; Erbring and Young, 1979). However, besides these

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

handfuls of very consequential empirical studies of diffusion, the bulk of the tradition in the area rests on the formulation of models that are testable with aggregate information about behaviors, such as the total numbers of adopters or proportions of a population who are adopters. These formulations mimic contagion models for the spread of disease and had the unfortunate consequence of portraying the social diffusion process as one where individuals are either passive carriers of information and innovations or passive “susceptibles,” rather than actors engaged in real interactions. Furthermore, these models almost always require the rather strong and frequently unacceptable assumption of temporal and spatial homogeneity. Stochastic versions of these early models make room for some types of heterogeneity, but have proved to be mathematically intractable and have stimulated little empirical research (Bartholomew, 1982; Bailey, 1975). New deterministic formulations of contagion models with explicit consideration of limited types of social heterogeneity have had little impact in sociological analysis (Anderson and May, 1992; for an exception, see Morris, 1993).

As shown in the review by Mahajan and Peterson (1985), conventional formulations of aggregate diffusion incorporate external sources and social interaction among individuals, and result in testable hypotheses about the progression of the number or proportion of adopters in the population over time. The classic representation with a logistics cumulative distribution (the “S-shaped” curve of adoption) is, in fact, a very general result, and holds up well under a number of formulations. The reasoning behind this formulation is that, if diffusion is mediated by interactions between individuals, it must be the case that the rate of change in the proportion P(t) of adopters is given by:

(1)

where A is the ultimate fraction of the population that will adopt, r0 is the number of new adopters that results from interaction with external forces, and r1 is the number of new adopters that results from interactions between adopters and potential adopters in a small interval of time, dt (the “diffusion yield” of social interaction). When r0 is 0 we have a simple case of pure social interaction effects, and when r1 is 0 we have a case of pure external effects and no social influence to speak of. This formulation does not distinguish between types of external sources nor between classes of social contacts as all interactions are considered the same, and all are thought to be equivalent in terms of their diffusion yield. Admittedly, one can complicate the formulation in a number of ways (see Mahajan and Peterson, 1985) to include the influence of several external sources and, more generally, to represent limited social heterogeneity.

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

For the most part, these modifications preserve the main advantage of the logistic process, simplicity, but do not supercede its main shortcomings, a result of the fact that the structure modeled is not complex enough to enable us to distinguish empirically among alternative processes. For these reasons, improvements in models that focus on aggregate outcomes are unlikely to generate significant progress. Thus, logistic and related aggregate representations of diffusion processes are increasingly relegated to the camp of fragile descriptions.6

About a decade or so ago, sociological analyses of diffusion moved in two different directions, away from conventional contagion models. The first and, as suggested above, perhaps least promising route was to reformulate logistic models to enrich the complexity of the structure being represented. The second and most promising was to reshape the object of study: rather than targeting aggregate parameters, such as the overall adoption rate, researchers began to focus on individual processes. This type of model shifts focus toward individual behaviors and individual adoption, and redirects attention to actors who are decision makers, to the processes of social influence that shape decision making, and, lastly, to the constraints to which these are subjected. The models eschew discourses about aggregate trajectories but formulate quite precise conditions for individual decision making that underlie a diffusion process. These models enable the researcher to fully incorporate complexities of the adoption process itself (interagent communication, external sources, barriers, agents of change, etc.) and the social conditions of interaction between actors who are adopters and potential adopters.

New Models for Aggregate Outcomes: Examples from Collective Action

An important step forward in the formulation of new diffusion models is the work on collective violence carried out by Pitcher et al. (1978). Their formulations were part of a more general effort to produce fruitful applications of diffusion models (Hamblin et al., 1973). The main notion behind their model is that the observed expression of collective violence depends on both imitation and inhibition processes determined by outcomes of prior events. Individuals learn from others’ behavior, including those participating in and those repressing violence, and are able to understand when and how collective violence occurs and what tactics seem to work best. As in all other aggregate diffusion models, however, it is the number of past events that determines decisions about adoption of violent behavior. Similarly, individuals are assumed to be homogeneous with respect to the relevant behavior (or characteristics determining the behavior), and events in the past influence current events in like manner (there

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

is time homogeneity of outcomes). With these simplifying assumptions, the authors formulate a model for the rate of change of acts of collective violence and the rate of change of repression acts. The model rests on two equations representing respectively the rate of new acts of violence and the rate of inhibition of acts of violence:

(2)

where P(t) is the cumulated number of acts of violence at time t and I(t) is the cumulated number of inhibited acts of violence. The algebra to solve for P(t) is transparent but tedious and results in the following function:

P(t)=P0 exp(δ/β) exp[(–δ/β) exp(–βt)] (3)

a model that represents the cumulated number of events (acts of violence), P(t), as a Gompertz’s distribution function. A Gompertz function is better suited to fit processes that lack the symmetry embedded in a logistic formulation, namely, those where the adoption process drags on through initially long and protracted stages before finally taking off. Asymmetry suits well most processes of collective violence studied by Pitcher et al. (1978). But although this curve fits the data better (see endnote 6), the most important innovation introduced here is that the aggregate model is derived from an ideal individual decision-making process whereby actors decide whether to participate in, abandon or avoid altogether acts of violence.

Modifications to the model introduced by Pitcher et al. (1978) that include an explicit definition of the growth process of repressive acts leads to an even better representation of the trajectory of collective violence (Myers, 1997). Not only does this formulation enable us to model the decision-making process of individuals who are potential adopters of the behavior but also the responses of those who are charged with the function of preventing those actions from occurring at all. The idea of formulating jointly two diffusion processes, one that fosters the behaviors of interest and one that inhibits their realization, should be of interest to those studying social process where the innovation, such as fertility control, may generate resistance on the part of central authorities or among influential members of the community (such as community elders, the church, provincial authorities, or even the state). This is, in fact, an elegant way to treat one element of the diffusion process, namely, the phenomenon whereby individuals cease to embrace or adopt a behavior. Yet,

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

although from the point of view of the theory of collective action this is much richer material, it continues to lead to a model for the aggregate number of collective acts of violence. And this is its main limitation.

A second type of diffusion models estimated with aggregate data relies on a more subtle representation of how individuals experience transitions from the state of nonadopter to the state of adopter (Rosero-Bixby, 1991; Rosero-Bixby and Casterline, 1993, 1994). Although the models are estimated from aggregate information (pooled cross-section and time series data on mean levels of fertility), their very nature (a close kin of compartment models) makes them suitable as representations of individual processes. Thus, the linkage between aggregate outcomes and individual behavior is more explicit here than what normally is in conventional diffusion models or even in the modified diffusion models for collective violence reviewed before. It is from this property that the model derives its superiority because it facilitates a richer formulation of the process than some of the models proposed by Pitcher and colleagues. The disadvantage of the compartment model formulation is that it is somewhat difficult to estimate from data normally available to us and, as other aggregate models of the same type, does not identify sufficient evidence to determine whether a diffusion process or something else explains the behavior under study (see, for example, simulations carried out in RoseroBixby and Casterline, 1993).

Models of Social Influence in Collective Action and Organization Theory

Somewhat paradoxically, an important part of the drift toward individually based models of diffusion occurs within areas traditionally reserved for the study of macrosocial processes, such as social movements and social organizations.

Collective Action

Initially, studies of common forms of collective action (e.g., protests and lynching), sprung from the idea that individual participation in such movements is a result of spontaneous and irrational imitation of antisocial behavior (LeBon, 1897). A contagion process was thus clearly justified as the best representation. This was replaced in the late 1960s and early 1970s by theories with an economic foundation that viewed collective action as the result of an atomized, individual decision-making process within a given social context and social environment (Olson, 1965). The overwhelming preoccupation in these formulations was centered in the so-called free-rider problem: to the extent that collectivities did not sup-

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

press the tendency of individuals to stay at the margins of actions, thus avoiding the costs of participation but reaping potential benefits, individual participation could simply be an irrational act. Reactions to Olson’s atomistic theory came from many camps but mainly from those who saw the absence of a role for social institutions and social interactions as a fatal flaw. Soon new theories were built around conceptual frameworks emphasizing economic and demographic conditions external to the movement or action (Olzak, 1992; McAdam, 1982); facilitation, repression, or channeling from the state or societal elites (Jenkins and Eckert, 1986; Tilly, 1978; Piven and Cloward, 1979; Pitcher et al., 1978); competition among protest groups (Tarrow, 1994); internal resources (Obershall, 1989); and the role of internal social processes and heterogeneity among actors or groups of actors that frustrate or promote the smooth organization of collective action (Marwell and Oliver, 1993; Myers, 1997).

Perhaps the most interesting developments in collective action theory take place with the introduction of the idea that the decision to participate in collective actions may depend to some extent on conditions associated with the individual’s position (e.g., individual costs or access to resources that facilitate action) and on the individual’s interpersonal relations. A sophisticated paradigm emerges, one that poses the existence of a diffusion-like or social influence process mediated by “the network structures of everyday life” (McAdam, 1995; cited in Strang and Soule, 1997:20). This change of focus is accompanied by a parallel displacement of the object of study: it is no longer sufficient or desirable to account solely for aggregate properties of the process (e.g., the proportion or number of protesters at a particular time or the rate of growth of protesters at the onset of the process). Instead, verification of richer theoretical specification of social influence requires modelling individual decisions and individual actions (Myers, 1997; McAdam, 1995; Valente, 1995; Laumann et al., 1977; Granovetter, 1973).

It is incorrect to think that recent theories in this area reduce the complicated processes that lead to collective action and determine its success or ultimate disappearance to diffusion or social influence processes. It is equally incorrect, though, to overlook the fact that it is in actors’ interactions and mutual social influence where one will find the essential features of collective action. This explains why, as indicated in the conclusion to a comprehensive overview of collective action theories and models, that “recent development in collective action models has centered on the problem of the interdependence of individuals within collectivities” (Marwell and Oliver, 1993:292). To be sure, there are other determinants and factors that should be examined, but without attention to social influence there is little hope of fully understanding how collective action develops.

Not surprisingly, modeling actor interdependency in collective action

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

requires concepts and tools that are suitable for modeling diffusion of behaviors. This is clearly evident in a number of recent models, from simple threshold models where actors’ decisions at one point in time are affected by the prevalence of participation or adoption among other relevant actors (Granovetter, 1973; Marwell and Oliver, 1993) to more complex constructions where an individual’s decision making evolves as a learning process or as a function of decision making among members of networks to which individuals belong (Marwell and Oliver, 1993; Laumann et al., 1977; Marsden, 1998). Threshold models, social learning models, and models of mutual influence are at the core of reformulation and representation of diffusion processes.

New theories of collective action (Coleman, 1990) rest on an assumption about an individual decision maker facing alternative action paths (adopting or not adopting a behavior also contemplated by other actors in the system). In doing so actors consider what others are doing. Who the relevant others may be and the exact influence they may exert on an individual’s behavior is possibly variable, and will depend on the actor’s position within the collectivity, his channels of communications, and the type and frequency of relations to others. It is at this juncture where the investigation of contextual effects and its connections to social networks becomes strategic for understanding collective action. Because these are also the foundations on which new diffusion theories and models rest, it is worth reviewing them in some detail. To do so I will begin from and then extend the Erbring and Young formulation of contextual processes.

According to Erbring and Young’s (1979) important contribution, contextual effects only make sense if they lead to the specification of a model where actors’ responses are a well-defined function of other actors’ responses. Assume, for example, one is studying a response for individual i, yi, and that we observe a vector of responses Y containing the values y1, y2,…yk, that is, all the information on responses for all relevant actors (1 through k) in the system. A proper model in Erbring and Young formulation requires that we define Y as a function of a transformed vector of responses:

Y=αWG(Y)+βX+∈ (4)

where Y is the observed vector of responses, ε is vector of errors, W is a matrix of weights, G is a well-defined functional form, X is a matrix of covariates, and β is a vector of associated effects. The central components in the model are W and α. The matrix of weights W, the “contiguity” matrix, specifies the importance attached to other actors’ responses by any one actor in the system. This matrix is what informs the nature of the network within which individuals participate, and the form in which their decision-making process influences all other members. The ith row

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

of the contiguity matrix contains elements that identify the weight that individual i assigns to the influence of the response of another member of the system. We could think of these quantities as measures of the degree of “infectiousness” of other members of the system (if they are infected) or social distance. In most cases one normalizes these quantities so the sum over all j is identical to 1 (Marsden and Friedkin, 1993). Clearly, the definition of W will vary depending on the mechanisms that generate or govern social influence. As they strive to understand achievement or aspirations among students of various classes in a school, Erbring and Young (1979:411) express the nature of the dependency of W on various social processes: “In the case of a contagion process, contiguity may be based explicitly on the amount of face-to-face interaction specific to each pair of students; whereas in the case of comparison or competition processes, contiguity may be defined as fixed and equal for all pairs of students in a given class room and zero for all pairs of students of different class rooms.” In the following examples, W is defined in a number of distinct ways.

The parameter a reflects the strength of the feedback from the group or collectivity. If this parameter drifts to 0, it is an indication that there is no social influence process affecting actors’ responses, and that these are only a function of structural characteristics (contained in X). One could generalize the formulation above by converting α into a vector, so that each individual in the system reacts differently to social influence and adoption by other members. In such case, the elements of the vector are equivalent to what individual infectiousness would be in contagion models.

As a consequence of this formulation, a necessary condition to prove the presence of diffusion of responses within the collectivity is that a (scalar or vector) be significantly different from zero. To the extent that W is misspecified, however, the estimates of a will be biased and incorrect inferences about social influences will be drawn. Thus, our ability to identify processes of social influence rests heavily on a proper specification of W. It is the task of general social network theories to specify what the nature of W ought to be, and what modifications we must introduce in (4) to capture better the social context which it is intending to represent. It is telling that social network theorists utilize formulations that are analogous, identical, or simple extensions of those proposed by Erbring and Young (see Valente, 1995; Marsden, 1998). I will show later that researchers in demography have also turned to variants of (4) to test new diffusion models for understanding fertility decline.

Note that model (4) is very flexible and that a number of variants could be tested. For example, suppose that Y represents responses at some time t and that G is the identity function. This simply means there is

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

a contemporaneous social influence. But if G is a lag operator so that G(Y) represents a vector of lagged responses, the model suggests that the process of social influence requires some time to be triggered and to exert significant effects on individual behaviors. Another useful extension is one where we postulate different matrices W for each lagged form of the vector of responses to reflect the possibility that actors’ mutual influence varies over time. Thus, for example, we can define matrices W1 and W2 to be associated with vectors of responses of lags 1 and 2 respectively, Y(t–1) and Y(t–2). Similarly, while the response variable y can stand for a dichotomous indicator at time t (actor adopts a response or not at time t), it is probably more informative to follow a sequence of values for y over time. Rather than modeling an actor’s response directly, one could choose to model the actor’s risk of adopting the response at some time t, μ(t), as a function of a transformation of actual responses of other actors in the system, G(Y). Finally, we could expand (4) to make Y a function not just of whether or not other responses have occurred but also of their observed outcomes. Thus, if adoption of a behavior by an actor could be classified as leading to “success” or “failure,” we could augment model (4) as follows:

Y=αWG(Y)+δW’G’(O)+βX+ (5)

where W’ is a modified contiguity matrix, G’ is a modified functional transform, and O is the vector of outcomes associated with positive (adopt) and negative (does not adopt) responses. In this model, evidence of diffusion or social influence must be retrieved from the estimates of a as well as from δ. And, as before, α and δ need not be scalar quantities but could be vector valued.

In what follows I briefly discuss an example from collective action research that makes use of these reformulations, and where the key empirical test is designed to identify the magnitude and direction of effects of social influence. My objective here is only to highlight the adoption of model (4). The technical difficulties in estimating its parameters is a theme discussed later.

The last example of diffusion models in collective action is the spread of trade unions in Sweden (Hedström, 1994). In this example, spatial relations are determinants of networks and networks participation, and these are the main factors determining the outcome of a mobilization effort. The main thesis is that spatial contagious processes exerted a decisive influence in the growth of the Swedish union movement.

Hedström’s analysis starts from a critical review of Olson’s theory and his unilateral attention to the free-rider problem and consequent inattention to social networks that generate dependency between actors’ decision making. In the case of Swedish trade unions, the claim is that deci-

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

sions to join the union movement are influenced not only by individual characteristics (the structural conditions) but also by the nature of their real or potential interrelations. The latter are in turn a function of spatial contiguity.

Hedström formulates a model for the hazard of a first union in a particular district. This model depends on (a) district-specific (structural) factors that are likely to promote (inhibit) the formation of unions and (b) an indicator of the network exposure to union formation. Because the data are in a discrete (year) period, the author chooses to estimate a logit model, rather than a continuous time hazard model, of the following form:

ln(pit/(1–pit))=αt+∑kkXikt)+βZi,t–1

where pit is the probability that the first union will be formed in district i in the year (t,t+1), Xikt is a characteristic k in district i at the beginning of year t, and Zi,t–1 is the weighted sum of union members in other districts in the year before t. The weights chosen represent the inverse of the distance between district i and all others. Note that with these weights, the variable Zi,t–1 is a simple function of the product of a contiguity matrix and a matrix of lagged responses in other units or districts. Indeed:

Zi,t–1=∑jπijnjt–1

where πij’s are the reciprocal of the distance between district i and district j and njt–1 is the total sum of union members in district j during the year (t–1,t). Thus this model has the classical form of other models for the spread of collective action. The evidence for (against) the existence of a diffusion process depends on the sign and magnitude of β. Hedström finds strong evidence that the onset of a first union is dependent on the spatial networks even when other factors accounting for structural conditions and national trends are considered. His conclusion is “the spread of information through the social or geographic landscape was of decisive importance for the formation of trade unions” and “the spread of the Swedish union movement was caused by a combination of local factors operating within districts and a contagious process operating between districts” (Hedström, 1994:1176). Or, translated in our jargon, the emergence of the Swedish trade union movements owes to both structural conditions as well as to diffusion processes.

Organizations

The formulation of diffusion or contagion-like processes and their application in organizational analysis is relatively new. Its most explicit and fullest development takes place within the so-called new institution-

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

alism framework. In what follows I discuss central elements in this framework and identify exactly where and for what purpose diffusion-like processes are invoked and utilized. I will attempt to show that these formulations are amenable to an analytic treatment very similar to those in collective action and that, as these do, they permit identification of all the definitional elements of a diffusion process.

Modern theorists of organizations have long been intrigued by the diversity of organizations and preferentially sought to explain heterogeneity in organizational structure and behaviors. Yet, proponents of the new institutionalism reverse the question and ask instead about the startling homogeneity in organizational forms and behaviors. The latter position is, of course, a revisionist version of the classic Weberian bureaucratic perspective that seeks to explain organizational uniformity by invoking the need to adopt rationalization to stay competitive and efficient. In some analyses of organization survival, the demise or failure of organizations is seen as a result of a selection process that weeds out the least competitive and efficient forms (Hannan and Freeman, 1977). According to the new institutionalism, this is only partially correct. Indeed, homogeneity in the organizational field is a result of two processes, one of which is driven by mechanisms of selection and competition (survival of the fittest or competitive isomorphism), whereas the other is one of institutional isomorphism. Its most distinctive characteristic is to be a result or consequence of adjustment in an environment populated by other organizations.

In an influential paper, DiMaggio and Powell argue that competitive and institutional isomorphism are applicable in general but that different fields of organization may be more or less prone to one or the other of these two processes. Thus, they suggest that “[competitive isomorphism] is most relevant for those fields in which free and open competition exists. It explains parts of the processes of bureaucratization that Weber observed, and may apply to early adoption of innovation, but it does not present a fully adequate picture of the modern world of organizations…” As Aldrich (1979:265) has argued, “The major factors that organizations must take into account are other organizations” (DiMaggio and Powell, 1991:62).

There are three mechanisms of institutional isomorphism: coercive isomorphism, normative, and mimetic. It is only the last that involves processes of social influence whereby organizations act and react to each other by adopting (rejecting) organizational features and behaviors, much as individuals are assumed to do in models of social contagion applied to collective action. Organizational mimicry could be construed as a diffusion-like process where the actual actors are not individuals but organizations themselves or key units within an organization.

The fundamental factor driving mimetic process is uncertainty. Ac-

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

cording to DiMaggio and Powell (1991:69), “when organizational technologies are poorly understood (March and Olsen, 1976), when goals are ambiguous, or when the environment creates uncertainty, organizations may model themselves on other organizations.” Just as in the case of individual actors, there are organizations that innovate and others that follow and imitate. Innovations are sometimes the result of imperfect imitations by one or more organizations of features observable in another that result in a modified feature that turns out to be beneficial for organizations in a particular field. As in individual processes of social influence, organizations are more likely to imitate those organizations in the field perceived to be legitimate or successful.

Isomorphism attributable to mimetic process is more likely to occur under a variety of conditions characterizing the organizational field or the organizations themselves. Thus, for example, to the extent that the connections between means and ends of an organization are unclear or uncertain, the more likely these organizations will be to adopt behaviors or features from other organizations in the field. Similarly, organizations with ambiguous goals will tend to imitate successful organizations in the field. Imitation can also be the result of threshold effects in the sense that adoption proceeds at a faster rate once the total prevalence of the feature exceeds a threshold value.

The propositions of interest with regard to organizational mimicry are very similar to the case of collective action, and the new models introduced to falsify them are, not surprisingly, also very similar. I will now illustrate these parallelisms of propositions and models with an example drawn from the recent literature in organizations.

The multidivisional form is a decentralized management organizational structure overwhelmingly preferred by those large firms that dominate the U.S. economy. Under this organizational form, “firms are organized into product divisions and each division contains a unitary structure. There also exists a central office where the long-range planning and financial allocations are located” (Fligstein, 1985:378). This organizational form could be considered a central feature of firms within an industry that adopt it. An interesting question is the following: what are the mechanisms that lead to the “spread” of this organization feature? Is it simple adaptation to conditions set by the U.S. market (of goods and employment), transportation technology, and the legal environment, or are there also imitation processes that trigger adoption of the form? This question was posed by Fligstein in an important article about ten years ago. To be fair, his effort was much broader, for he attempted to discriminate between several alternative theories, all of which could account for the multidivisional form, but only one of them involves a mimicking process.

Fligstein explicitly models the adoption of the organizational feature

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

with simple logit models, each of which is defined for one decade during the entire period under observation (1929–1979). Each logit model is for a dependent dichotomous variable that indicates whether the feature in a firm is adopted or not. Several independent variables capture essential features of the mechanisms postulated by competing theories (structural factors). He then includes in the model the percentage of firms in a given industry that adopted the feature before the beginning of the decade under observation. This is the variable that represents the crucial feature of a mimicry mechanism. For each period we have (t, t+10):

ln (pt/(1–pt))=βtXt

where βt and Xt are vectors of covariates and effects for the period.

Although in a very simplified form, the model used by Fligstein is an example of the relational model for social influences introduced before. Indeed, the indicator of prevalence of the organizational form in a given industry is a summary indicator of the information contained in a lagged response matrix (Y) combined with the identity matrix as a contiguity matrix.

Fligstein’s findings suggest that there is evidence indicating that mimicry does operate in the transmission of the multidivisional form. This evidence is not overwhelming as there is also support for the existence of other mechanisms of isomorphism. Inferences about the existence of a diffusion process, however, are somewhat weak, not just because the evidence is less strong than desirable, but also for two other reasons. First, as formulated the model cannot identify exactly how the imitation process proceeds, that is, it does not shed light on the micromechanisms (at the level of sections or units or single individuals in a firm) that sustain the imitation process. Second, there are a number of statistical problems that the author cannot resolve with the data available to him, and they all involve issues of proper (inconsistent) estimation of parameters. These will be reviewed more thoroughly in the next section. Despite these shortcomings, however, Fligstein’s work represents the first attempt to explicitly test DiMaggio and Powell’s mimetic mechanism.

NEW MODELS OF DIFFUSION: PROBLEMS AND UNRESOLVED ISSUES

The discussion and review of recent sociological applications above provide elements for identifying essential characteristics of diffusion models and for testing propositions that seek to identify their relevance in empirical cases. Unlike conventional diffusion models, the new models applied in sociological analysis formulate explicitly the mechanisms

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

through which diffusion occurs and provide an environment for testing their empirical relevance.

In this section I will identify the main characteristics of these models, establish the advantages gained by adopting them, note important shortcomings, and discuss possible improvements. Throughout, the discussion is focused on the following query: how can we empirically identify a diffusion model, that is, how can we tell it apart from altogether different mechanisms? This discussion will furnish a “golden” standard that will be used later to assess diffusion models applied to demographic problems.

A Simple Representation of Diffusion Processes

In light of our previous discussion, we introduce a modified version of the definition given earlier. A diffusion process is one in which selection or adoption (rejection) of a behavior or practice depends on an individual decision-making process that assigns significant influence to the adoption (rejection) behavior of other individuals within the social system. There are a number of ways to define who the other individuals are, and there are alternative mechanisms through which their social influence may affect an individual’s decision-making process. In what follows I briefly identify the most significant social relations and three mechanisms that drive diffusion processes.

I start from a simplified version of the decision-making process worked out by Montgomery and Chung (1994) (see also Montgomery and Casterline, 1996) and assume that we are dealing with the adoption (rejection) of behavior Bo and that individuals may choose among a repertoire of alternative behaviors contained in the set {Bj}, of which Bo is a member. Each of these behaviors is associated with expected costs and expected benefits. Assume that individuals associate each behavior B, with a distribution of net benefits, NBj. Let us say for simplicity that NBj. is continuous, can attain values in the interval (–ñ ,+ñ), and is associated with probabilities Pj(x), where x is a given level of net benefits. Each individual assigned to behavior Bj receives a net benefit, NBj=x with probability Pj(x). This is what we will refer to as linkage between behavior and net benefit. That is to say, for any behavior Bj. there is an expected net benefit given by:

E(NBj)=∫x(NBj (x))Pj(x)dx.

The decision-making problem is simply to choose the behavior within the set {Bj} that maximizes E(.).

In the absence of a diffusion process, the inclusion (exclusion) of Bo from the set of alternative behaviors, the actual configuration of the set of equivalent behaviors itself, and the probabilistic association of net

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

benefits depends on the actors’ position in the social system or, rather, on the bundle of resources (including information) associated with or available to the actor. This is what a structural explanation points to: the selection of behavior is solely dependent on characteristics associated with the individual, not with what others do regardless of whom they may be. Instead, a diffusion process exists when either the inclusion of Bo in the set of alternative behaviors, the linkage between behavior and net benefits, or the actual calculus of costs and benefits depends also on conditions dictated by social contacts with other members of the system, however tenuous or formal these may be. That is, these social contacts or social influences are effective mechanisms of diffusion in that they have an effect on (a) information about the feasibility of Bo, (b) knowledge about net benefits associated with Bo, or (c) assessment of net benefits associated with Bo given a nonzero prevalence of Bo in the social system.

To make the above definition unequivocal, a number of issues require clarification. First, we need precision in the timing of the individual calculus. Thus, we need to know the time horizon for the calculation of net benefits and, more importantly, the time lags required to establish an association between a behavior and its net benefits. If there is no diffusion at all, the time lag may be instantaneous, very short, or quite long, but if there is diffusion individuals will surely require some time to learn from others’ experiences about rewards and costs associated with Bo. And if this is the case, how long does it take for the association to become established from observation of others’ behavior?

Second, it was assumed that decision makers are only interested in the mean of the distribution of net benefits, and that issues such as higher risks imposed by higher variances are irrelevant. This assumption may be inadequate both when there is diffusion and in the absence of it, but more so in the first case (Montgomery and Chung, 1994). In fact, when there is diffusion and individuals purposely take into account others’ behavior, it is likely they will have only sparse information on rewards and costs of adoption of Bo, particularly at the onset of the process. In such cases, the distribution of net benefits will have higher variances, and risk-averse individuals will have a harder time adopting the behavior, regardless of what the mean net benefit is. This may be one explanation for the phenomenon of resistance to adopt, which is especially relevant at the onset of the diffusion process. We elaborate on this below.

Third, to identify the diffusion process from observables, one needs to know with precision about which social networks are relevant and which relations are key within them. This is the material informed by network theory; it is discussed briefly below. In addition, we should consider two additional difficult issues. First, the formation of an individual’s

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

reference networks may be endogenous to the process being studied. This means that the selection of social networks or of relevant relations within them could be influenced by the same factors that affect decision making about adoption of a behavior. For example, suppose that individuals in a given social position tend to choose a behavior Bo based on maximization of net benefits purely as a function of their position in the social system, and that there is no influence of others’ behavior in their decision-making process. If, to avoid social friction, social rejection, or complete isolation, they decide to choose social networks (and relations within them) whose members have also chosen Bo, the empirical process will appear as though individual adoption of Bo was associated with prevalence of Bo in relevant networks. The incorrect inference is that there is a diffusion process because the probability of adoption of the behavior will be associated with the relevance of Bo in the individual’s network. But in this example membership in a network follows adoption of behavior, not the other way around. The only way to avoid an incorrect inference is to have full information about the timing of the adoption and the timing of effective membership in networks.

The second and related issue is of great relevance for current processes in this volume of diffusion of contraceptive (and other) behaviors via the influence of television (Potter et al., 1998). As suggested by Montgomery and Chung (1994), television creates fictional networks with which individuals identify and participate. The television program communicates the existence of alternative behaviors (plausibility of contraception, for example), but also transmits information about the connection, usually spurious, between the behavior and desirable rewards. Thus, admired couples in soap operas may have no more than two children, live in a mansion, and drive red Ferraris. To the extent that these are desirable objects, they will be associated with two children at most. This may introduce, reaffirm, or consolidate the idea that children are costly. Thus, although television is an external source, it can operate much in the same way as membership in social networks does and, therefore, raises the same issues of selectivity alluded to before.

Fourth and last, the preconditions for the existence of a diffusion process stated above refer to mechanisms through which social networks affect individual choice of behavior. Montgomery and Chung (1994) suggest that there are two mechanisms: one is by altering knowledge about the elements of the set of plausible behavior and the other is by establishing a linkage between the behavior and its net benefits. There is a third mechanism through which costs and benefits associated with adoption are changed by the diffusion process itself. An example that shows this type of effect is the spread of a technological innovation, such as an operating system. The adoption of the innovation changes the conditions for

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

everybody else, whether they are adopters or not. For example, adoption of an operating system among some users prompts the creation of software designed for the operating system, but the software may be useful to the entire collectivity of PC users. The immediate effect should be to increase the likelihood of adopting the operating system simply as a way to access the software. Examples such as these are easy to identify in the area of technology, and they are not altogether absent in the area of social behaviors. Thus, the social costs of refusal to participate in collective action may grow steeper as members of a reference group increasingly adopt the new technology. Similarly, the adoption of contraceptive behavior may become more plausible if social and economic conditions emerging after and as a result of the initial adoption of contraception impose steep costs among large families.

This third mechanism acts by speeding the spread of the behavior, but is the result of a feedback and operates through adjustments that individuals make to changes in costs or benefits, not as direct response to others’ adoption. It is a mechanism that augments (inhibits) the diffusion or spread of the behavior by activating the structural factors that affect behaviors. The feedback mechanism operates through influence of others’ adoption on costs and benefits associated with known behaviors, not by altering awareness about a set of options nor by establishing a new connection between behaviors and net benefits. Furthermore, the influence of the feedback effect on an individual’s adoption may be exerted by diffuse and distant social networks, not necessarily by any specific social network to which the individual belongs.7

Below I describe in more detail some of the problems we encounter in the definition and treatment of relevant social influences and individual resistance and thresholds. I then discuss considerations for model building.

Relational and Structural Models

The most recent diffusion models that explicitly incorporate the effects of social network do not neglect the existence of traditional elements altogether. Insistence on the influence of external sources remains an important feature, and the new formulations may even include nontraditional external sources (e.g., those that regulate the environment within which decisions are being made). What is novel in these models is a more detailed treatment of the mechanisms through which external sources affect the adoption process. It becomes relevant, for example, to formulate precisely whether a televison show or a particular radio program affects values or preferences, whether it facilitates communication of information among individuals in different social positions, or whether it alters

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

the costs of certain alternatives. The task of model formulation becomes a more taxing one because we must hypothesize in advance the mechanisms through which external sources are thought to affect the adoption process.

The mainstay of these models is attention to the sources of social influence and the attempt to model these as precisely as possible. Classical formal models of diffusion assume spatial and social homogeneity, that is, they rest on the assumption that members of the population do not differ in terms of the chances of affecting others or being affected by others. Sociologists now postulate that there are at least four different mechanisms of interpersonal relations that shape the social structure within which adoption decisions are made. Each of them requires a different modeling strategy. First, relational linkages refer to the set of relations an individual may establish with others within a particular setting or network. What matters here is the density of individual connections, as well as the type of connections that some actors in the network have with others outside it. These relations can be represented by an individual, vector-valued function. Each individual is characterized by a vector-valued function, the contiguity matrix introduced above, that reflects all social connections considered to be relevant. To the extent that relations maintained by others in the network with individuals outside it are relevant for the process, they can be incorporated into the matrix in the form of weights. For example, a set of weights might distinguish the relative importance of strong and weak ties for a given individual in the population (Granovetter, 1973).

Second, structural linkages refer to relations with structurally equivalent actors. More generally, they are relations between individuals evaluated as a function of similarity of structural positions occupied within a given network or in the wider social system (Burt, 1987). Application of this idea is fairly common in recent studies of organizational diffusion. Thus, it is thought that structural equivalent relations promote imitation through competition among individuals in a firm. But it can also be the case that competition with structurally equivalent actors may spur not imitation but divergence of behavior (that is, resistance to adoption). This is an empirical matter and can be settled only if we are able to associate with each individual a matrix-valued function of relations to individuals who are structurally equivalent (or dissimilar). The empirical estimates associated with such matrix-valued function will enable us to determine the direction and magnitude of effects.

Third, the new models incorporate and account explicitly for the degree of influence accorded to others within a network. Influence attributable to a position or member will usually be a function of the relative

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

ordering of individuals in the network according to some relevant metric (Valente, 1995).

Fourth, the models include consideration of influence exerted by culturally bounded groups (Strang and Soule, 1997). These refer to relations maintained with individuals based on definitions of actions, status, and purpose. For example, the influence of individuals who consider themselves as activists (McAdam and Rucht, 1993) was singled out as an important factor in the spread of activists’ tactics. An interesting but somewhat puzzling type of cultural influence might be one generated by the innovation itself. Thus, individuals may align themselves around the notion of being or not being adopters. This could influence continuation of adoption and attract (or even repel) nonadopters.

Fifth, and finally, spatial proximity can be incorporated into the analysis. A common finding in classic diffusion research is that spatially proximate actors are more likely to influence each other. The difficulty is that spatial proximity is an open concept in the sense that many mechanisms can operate to render it an effective means to promote (or resist) adoption. In most cases, spatial proximity is used as a proxy, albeit imperfect, of network connections and potential social influences originated in either structural, relational, or cultural connections as defined before.8 Ideally, the use of spatial proximity should be justified a priori by defining the precise mechanisms through which it may affect the process. Some of these mechanisms are easiness of communication, social and economic homogeneity, and frequency of interactions.

For the most part, the study of effects of spatial proximity has been monopolized by geographers (Brown, 1981), but the development of tools for statistical inference from spatial statistics (Cressie, 1991) and the rapid adoption of accessible software (Anselin, 1988) has promoted the use of spatial models of diffusion (see, for example, Hedström, 1994; Bocquet-Appel, 1997).

Resistance and Thresholds

An important innovation introduced in recent formulations of diffusion models is the notion of individual thresholds. According to this idea, individuals may resist adoption up to a certain point as the process proceeds within the group. In theory at least, individuals may be carriers of different thresholds and thus be characterized by a unique, individual-specific value identifying the percent of total adoption below which efficient individual resistance to adoption will be exerted. This is, of course, an unmeasured quantity, in much the same way as frailty in the literature on health and mortality is an unmeasured quantity. If all variables affecting the individual decision-making process were known,

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

it would be pointless to speak of thresholds, in much the same way as the notion of individual frailty would be empty if all conditions for survival were known. For example, one mechanism promoting (slowing) adoption that requires the notion of individual threshold also requires invoking the notions of variance of the distribution of net benefits associated with the behavior to be adopted (see above) and of risk aversion. A high variance of net rewards will prompt risk-averse individuals to delay or reject adoption. In this interpretation, individual resistance and thresholds are a function of (a) perceived variance of net benefits and (b) whether or nor an individual is risk averse. Neither of these are readily measurable quantities.

Initially at least, proponents of the notion of threshold have made it equivalent to the effect that levels of prevalence of adoption within the relevant group has on the individual risk of adopting (Valente, 1995). But this is a conceptually different idea and it creates relatively serious identification (Manski, 1995) and interpretational (Erbring and Young, 1979) issues that have not yet been addressed satisfactorily. Because the notion is quite promising, however, it is hoped that work toward the development of better measurement and modeling strategies will continue.

SEARCHING FOR DIFFUSION: THE IDENTIFICATION PROBLEM

The Ideal Test

The only way to conclusively prove whether a diffusion or a structuralist theory is correct is an unrealizable experiment, namely, the observation of patterns of behavior under conditions that hold constant the distribution of individuals by social positions and the distribution of resources associated with positions while allowing variations in conditions that trigger the spread of the behavior (e.g., participation in social networks). If the prevalence of the behavior grows, it cannot possibly be because of structural factors (they are being kept constant) but because of diffusion. The key issue, however, is to remember that at least one of the three mechanisms of diffusion identified above mimics the effects of structural changes, namely, when social positions or resources associated with them change as a result of the process of diffusion itself. Put otherwise, if we are to identify diffusion effects, the ideal experiment cannot allow the diffusion feedback mechanism to operate and simultaneously maintain invariance in individual characteristics. Thus, even under ideal conditions, it is difficult to sort out precisely how much of the ultimate change in behavior is due to all diffusion mechanisms and how much to secondary changes in the social structure induced by diffusion itself. In the case of the study of fertility or of the bulk of social sciences problems, where conventional

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

study conditions are far from ideal, it will be virtually impossible to make the relevant distinctions. This limitation is, of course, irrelevant when the feedback mechanism is weak or if its operation requires long time lags.9

Testable Models

The recent literature in diffusion models is very rich in suggested formal representations (Strang, 1991; Strang and Tuma, 1993). I will borrow freely from these but tailor the discussion to capture useful features for demographic analyses. My purpose is not to suggest what the true models are, but rather to provide an indication of the degree of complexity that the models ought to have, and to point to the problems one is likely to face when a diffusion model is misspecified.

A diffusion model must represent individuals choosing among a set of alternative behaviors under a set of constraints. It must also account for the persistence of the adoption or selection over time. This can be done in a number of ways, but perhaps the most efficient one is to construct a system of two states, one representing adoption of the target behavior and the other adoption of a different behavior (or refusal to adopt). Individuals may move between these two states as a function of individual characteristics associated with social and economic conditions (costs and utilities), external characteristics acting as constraints (or facilitators), influence of external sources of ideas, or influence of individual social networks. To capture the dynamic of this two-state system, we can formulate a pair of equations for the risk or hazard of transitions between the two states:

(6)

where μ12i(t) refers to the risk of moving from state 1 (nonadopter) to state 2 (adopter) for individual i at time t, μo12(t) is a baseline hazard, Xi is a vector of structural characteristics of individual i, Zi is a vector-valued function containing information on external sources of information that may influence ith’s choice, Wi is a contiguity vector for individual i containing the weight assigned to the influence from contacts with individuals j=1,…N, where is N is the total number of members in the system, G is a functional transform, and Y is a vector of responses for members j=1, …N. Finally, ε12i is an error term. The second equation defines the risk of moving from state 2 to state 1 (abandoning the new behavior). It is analogous to the first, but I have allowed for the possibility of different baselines, different effects, and different matrices of covariates. The contiguity vector is time dependent to allow for changing influences derived

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

from social networks during the process, and so are the vectors of responses Y and Y* to allow for updating of information about members of the system.

Problems with Formulation of the Model

Before reviewing estimation problems, let us examine the anatomy and functionality of this formulation. Suppose this two-state model is correct. Under fairly general regularity conditions, there will be a steady state and a stable proportion of the population that will be in each of the two states. It is not difficult to show that those proportions or their logarithms are NOT a simple linear function of the vector of covariates, as the logarithms of the risks are. This statement is important: it means that if model (6) is the true representation of the diffusion process, aggregate linear models for quantities such as the proportion of adopters are misspecified. In addition because the model is misspecified, it is totally meaningless to estimate its parameters and to attribute to diffusion the part of the variance in the dependent variables (proportion of adopters) that remains unexplained by measured covariates.

Furthermore, let us say that variables are scaled in such a way that α and γ are positive. It follows that if diffusion is effective, the adoption process will proceed faster than it would otherwise (the risk of adoption will be higher and the probability of staying in state 1 will be lower). But this does not mean that one is correct in inferring the existence of diffusion if the change in the aggregate proportion of adopters is “rapid” or “fast” (relative to some standard). This is because (a) we assume that the second transition is nonexistent and (b) we assume that all relevant structural covariates are contained in X. Even if the second transition was irrelevant (all adopters remain adopters for life and beyond), lack of appropriate control for structural conditions that change rapidly and that have strong effects on the risk of adoption will end up concealing the extent to which the processes is structurally driven and mislead the investigator into believing that the whole process is the work of diffusion. Note that this will occur even if one is estimating model (6) rather than an aggregate variant of it, regardless of whether X’s are unrelated to Z’s.

A number of difficulties are associated with any possible extensions of model (6). First, we have not justified well the nature of the term associated with social networks. In particular, there is no reason why it should include all members of an individual network. An alternative representation would be to split the term into two components in the following way:

αWi(t)G(Y(t))=α1W1i(t)G1(Y1(t))+α2W2i(t)G2(Y2(t))

where each vector now refers to individuals in the network relevant for

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

the ith individual according to their response, and collects all those who have not adopted in W1i and all those who adopted in W2i. This partition will enable us to distinguish the effects of attraction toward the new behavior (exerted by those who adopted the behavior) from the potential resistance effects exerted by those who have not yet adopted.

A second problem is whether the model must be multiplicative at all, that is, one where the diffusion component and the structural component enter as multiplicative terms. Strang and Tuma suggest pursuing an additive model because this has some desirable properties (Strang and Tuma, 1993). Because estimates and implications associated with each of these formulations will be different, the researcher must think through the assumptions made when adopting one or the other form.

Third, effects of diffusion operating through external sources should be captured by γ, whereas effects of social influence will be captured by α. Both γ and α can be allowed to be vector valued, that is, individuals may have different susceptibility to be influenced by others, depending on who the other individual is. If one considers this an important extension, then problems of identification will emerge. Whether α and γ are scalar quantities or vector-valued functions, their magnitude and sign will only reflect two mechanisms of diffusion: one whereby social influences change the set of plausible choices for the individual, and the other whereby social influences modify the linkage between the new behavior and expected net benefits. These effects will not capture the influence of the diffusion process via the feedback mechanism. To capture the feedback mechanism, model (6) must be extended to reveal the relation between prevalence of the new behavior and structural conditions contained in Xi. Alternatively, if the feedback mechanism requires a long time to operate relative to the speed of diffusion to be significant, one could simply dismiss it.

Fourth, there is no need to have a unique contiguity vector, Wi. In fact, one could partition the vector to reflect several (partially related) networks or to attempt to represent functional and structural influences (see discussion above). Furthermore, one could introduce a vector representing the success associated with the adoption of the behaviors by members of relevant networks (see above). These two modifications increase the richness of the social network representation, but they also pose additional data demands.

Problems of Estimation

If we insist on the existence of two transitions, the first problem that emerges is that of the relation between the two error terms. Without assuming a joint distribution for the errors (and, inevitably, this will be

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

arbitrary), the parameters are not estimable. Of course, the easiest but least appropriate solution is always on hand, namely, to assume that the two error terms are independent.

But even a simplified, one-equation form of model (6) creates estimation problems of considerable import. Assume the simplified form that results when the transition from state 2 to state 1 (abandoning adoption) is insignificant. That is, we assume that once adoption occurs, it is irreversible. This is consistent with the process of fertility decline in general, although it may not be with other diffusion processes or with some examples of local fertility decline. Also, to simplify even further, assume there is no relevant feedback mechanism. The parameters of the simplified model that remain will be estimable only if we have available considerable amounts of information, and if we make some strong assumptions along the way.

The required information includes (a) nature of several types of networks relevant to all individuals in the population during the trajectory of the process, (b) information on outcomes associated with adoption of the new behaviors by members of the networks in which all individuals participate, (c) nature of the sources of external influence to which each individual is susceptible throughout the duration of the process, and, finally, (d) structural conditions that determine either individual positions or external constraints to take into account in individual decision making. Needless to say, there are very few datasets in sociology or demography that contain all this information, and even fewer social scientists who will be able to discern what all the relevant variables are. As a result the researcher faces the problem of unmeasured heterogeneity whereby estimated effects are inconsistent even if the omitted variables are unrelated to the included variables. It is not difficult to design scenarios where omission of a structural characteristic could impart an upward bias on some estimates thus misleading the researcher into believing there is a nontrivial diffusion process. Note that, unlike most cases where generalized linear models apply, the biases or inconsistencies will occur regardless of whether the omitted variable is related to those included in the model. Earlier we identified one potential culprit of unmeasured heterogeneity, namely, information on the appraisal of the risk to which an individual is exposed when connecting the new behavior to net benefits. If risk-averse individuals perceive a larger variance than others, they are likely to delay adoption. This will lead to a well-known artifact: the risk of adoption will look like a decreasing function of time. The most likely consequence of this will be to bias downward the effects of diffusion.

Unmeasured heterogeneity can be modeled, and one of the most effective ways of doing so is to postulate that each individual is characterized by a resistance (or susceptibility) to innovation or a “threshold” for

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

innovation. This unmeasured individual characteristic is postulated to be a random variable with a known distribution. The assumption just made leads to calculations that result in a marginal risk that is not dependent on the unmeasured characteristic. Thus, the formulations of diffusion process or collective actions that invoke the existence of individual resistance or individual thresholds are really designed to interpret the existence of unmeasured characteristics that either promote or delay the adoption of the behavior. The hazard model formulation offers a framework showing where to include them.

Finally, a more troublesome feature of a diffusion process is that its own progress may affect the likelihood of reducing, eliminating, or inventing new social networks. Maintaining social networks that are not responsive to the new behavior may force adventurous individuals to seek new social attachments among those better prepared to embrace the new behavior. Although this avoids friction and possible penalties, it is associated with new costs, and individuals will need to weigh the advantages of leaving current social networks perceived to be unfavorable against the cost of creating new relations in newer and more receptive networks. Ultimately, however, what matters is that such endogenous change will produce the appearance that networks do have an influence on choice of behavior when actually they may have none. Naturally, the only way to handle this problem is to model separately network formation as a function of past behavior. This imposes more information constraints and generates new estimation difficulties.10

DIFFUSION MODELS THAT ACCOUNT FOR FERTILITY CHANGES

The history of applications of diffusion models to the study of fertility proceeds through a succession of intellectual stages, each characterized by a paradigm. These paradigms serve to conceptualize the nature of the process of diffusion, to identify the criteria for deciding empirically among competing explanations, and, finally, to define the opposition between two types of explanations for fertility decline, one relying on diffusion and the other on structural changes. The initial dominant paradigm is characterized by a naive conception of diffusion, undemanding methodological desiderata, and a simplistic contrast between diffusion processes and structural changes. This paradigm gradually gives way to more sophisticated and subtle views that include models closer to the ideal set forth above, adhere to strict principles of empirical inferences, and represent better the contrast between diffusion and alternative explanations in a more elaborated form. The limitation faced by this more recent para-

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

digm is not conceptual but empirical, for its inferential principles require more information than what is normally available to us.

The Origins of the Contrast Between Structuralist and Diffusionist Explanations

The dichotomy between structuralist and diffusionist explanations of fertility change was first formulated and translated into testable hypotheses in a seminal paper by Carlsson (1966). In it the author establishes a rather misleading contrast between processes of adjustment, rather than structural transformations and diffusion. Carlsson argues that fertility decline in Sweden could have been triggered by one of two processes. The first was an adjustment process whereby families tinkered heavily with their fertility targets to accommodate higher levels of child survival. Alternatively, fertility changes could have been the result of a diffusion process whereby a few forerunners introduce controlled fertility, and were subsequently imitated by others who adopted the innovation. The adjustment hypothesis is a corollary of a classic formulation of demographic transition theory that assigns importance to mortality decline as a precursor to fertility decline (Davis, 1963; Notestein, 1945; Knodel and van de Walle, 1967).

Carlsson’s dichotomy is misleading for two reasons. The first is that the adjustment hypothesis is indeed one possible version of a structuralist argument, and in presenting it as a sharply different, competing explanation to a diffusion mechanism, it falls into the trap that has had us confused for years. The confusion is that we reify diffusion as a process that involves at best a-rational individuals, whereas the structurally based explanation invokes rational decision makers. I have pointed out that this interpretation is incorrect, and that the difference between one and the other has nothing to do with actors’ rationality, and everything to do with the existence of two distinct rationalities: one where others’ opinions and actions count for fertility decisions, and another where they exert no influence on the calculus of fertility.

Carlsson’s distinction is misleading for another reason as well. This is that the main mechanism through which the process of adjustment is assumed to work involves improvements in child survival. If there is anything we learned from evidence gathered in Western Europe and in developing countries, it is that, in most regions of the world, mortality decline had little to do with fertility decline, and none of the three main mediating mechanisms linking mortality and fertility—biological, replacement, and hoarding—are powerful enough to amount to a full explanation (van de Walle, 1986; Preston, 1978; Cohen and Montgomery, 1998; Palloni and Rafalimanana, 1997). Few among us would argue that failure

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

of this particular (Carlsson’s) version of the structuralist argument is reason enough to sponsor the diffusion hypothesis.

Carlsson’s formal representation and testing of a diffusion model is fraught with other problems as well. The most important among these, and a consequence of the simplistic notion of diffusion adopted, is that there is no model connecting individual fertility decision-making and social influences. Although the author does refer to geographic proximity, there is no effort to include it explicitly in a model and it is, instead, introduced as an ad hoc variable.

Finally, Carlsson (1966:173) reduces diffusion to a simple flow of information from an external source to individuals. Indeed, he represents his work as an effort to induce a shift from “innovation to adjustment theory [which] leads to less emphasis on information about birth control and its means, and more emphasis on motivation and social situation” (emphasis added). As we showed before, this reduction of diffusion to supply of information is misguided. Interestingly, he anticipates potential reversals in the diffusion process when he recognizes the possibility of individuals rejecting the new behavior after it has been adopted, but draws the wrong conclusion from it. Indeed, he speculates “One reason why it may be more misleading than helpful to regard the fertility decline and the wider adoption of birth control as an innovation process is that the latter designation often carries with it the idea that the process is bound to run its course to complete or near-complete adoption in a regular way. The notion of an adjustment over time to a new equilibrium level, on the other hand, keeps open the possibility of fertility staying neither fully controlled in the modern sense, nor completely uncontrolled, and this for an appreciable period” (Carlsson, 1966:172). This quote suggests that despite identifying the problem of rejection (and reversals) as a central one, he does not conceive of a solution within the boundaries of a diffusion model, as we did earlier using a two-state hazard model.

The Princeton Fertility Project and Its Aftermath

There is widespread agreement that the results of the Princeton project cast doubts on the validity of classical explanations of fertility changes. These results led and encouraged formulations of more refined interpretations of the diffusionist models. By far the most damaging empirical evidence produced by the Princeton project against a paradigmatic version of the structural explanation, conventional demographic transition theory, is that fertility decline appears to occur along territorial boundaries reproducing ethnic, language, and religious cleavages (Lesthaeghe, 1977; Livi-Bacci, 1971, 1986; Knodel, 1974; Coale and Watkins, 1986). Conventional regression analyses reveal that the explanatory power of vari-

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

ables measuring industrialization, urbanization, state centralization, bureaucratization, and others to predict the onset and the pace of fertility decline turns out to be modest at best. This evidence led to a more refined representation of how diffusion and resistance to diffusion may operate in societies sharply divided along linguistic and ethnic cleavages. Yet, with the exception of work by Lesthaeghe on Belgium’s fertility decline (Lesthaeghe, 1977), the contrast between the structuralist and diffusion theories is always resolved by estimating conventional linear regression models on aggregate indicators of fertility, then resorting to residual analyses as a tool to assess the degree of failure of the structuralist theory. This failure is automatically considered as a sort of reverse measure of the degree of success of the diffusion model. Thus, although most studies in the Princeton project attempted to conceptualize more precisely the process of diffusion, adding the idea of cultural and ethnic boundary and refining the conceptualization and measurement of structural conditions, the rules of inference remained quite primitive.

The paradigm that characterizes the Princeton fertility project has been subsequently modified along three lines of research. The first introduces more fine-tuned analyses of the same or moderately augmented data used in the Princeton study without significantly changing the theoretical discourse (see the work by Galloway et al., 1994; Bocquet-Appel, 1997). The second line of research focuses on different measures of fertility, correctly arguing that the proper measures to test diffusion models ought to be measures of prevalence of the new behavior (contraception) that are only poorly correlated with the indirect measures of fertility normally used by demographers (Okun, 1994).

Finally, the third line of research is more theoretical because it refines the conceptual scheme and brings to the forefront the discussion of the nature of mechanisms whereby individual adoption of new behaviors takes place. This occurs in reaction to overwhelming evidence of the failure of conventional, structuralist explanations of fertility changes. At the end of the 1990s, demographers had already surveyed extensive territories in addition to Western and Eastern Europe. The World Fertility Survey, the Demographic and Health Survey, and a handful of other more localized data collection undertakings produced a large amount of evidence regarding fertility decline in developing nations. In a sweeping and controversial summary of this evidence, Cleland and Wilson (1987) suggest that any version of demand theories of fertility, that is, economic theories invoking the need for structural changes in individual’s positions as a precondition for fertility changes, cannot account for the onset, pace, and geographic location of fertility declines throughout the developing world. Instead, these changes appear to be driven by ideational changes riding on the back of a diffusion process. Much the

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

same conclusion had been reached by Caldwell in some of his writings where he assigns importance to the onslaught of an ideational change (“Westernization”) that precedes and is partially independent of changes in forms of production and distribution (Caldwell, 1982).

However persuasive their argumentation may be, the formulation put forward by Cleland and Wilson runs into two problems, one theoretical and the other methodological. First, there is a conceptual confusion that takes ideational changes as equivalent to diffusion processes. If fertility declines because individuals change ideas about the advantages of having children, even though their social and economic positions remain apparently the same, one cannot automatically infer the existence of a diffusion process. For this to be a proper inference, one must find evidence that the new ideas or the change of ideas are driven by imitation of others’ ideas. Second, the evidence that Cleland and Wilson use to support their claims belongs to a type we identified before as insufficient. Indeed, they examine the speed of changes in fertility and compare them with what would be expected given observed changes in structural conditions or, alternatively, they verify that the main cleavages created by fertility changes are drawn by ethnic or language distinctions. This contrast between ideational changes and demand-driven changes at the core of Cleland and Wilson and Caldwell’s formulations are reminiscent of the coarse contrast between adjustment and diffusion already contained in the older paradigm used by Carlsson. More recently, Bongaarts and Watkins (1996) review aggregate empirical evidence regarding the timing and pace of recent fertility declines. As Cleland and Wilson do, they too reach the conclusion that much of what we observe during the past twenty or thirty years is attributable to the transmission of information and ideas regarding fertility control. Their conceptualization of what is being transmitted and how it is transmitted is broader and perhaps more precise than Cleland and Wilson’s because it includes both micro-level diffusion processes (at the level of local networks and peers) as well as macro-level diffusion (global and national networks). But their inferences are based on linear shifts analysis, a device that rests on the unverified assumption that the magnitude of unexplained variance accounted for by shifts is associated with mechanisms facilitating diffusion. This may be suggestive but it is not the kind of proof we require to verify the existence of diffusion processes.

Robust Theoretical Formulations

The paradigm that characterizes the third stage in the history of application of diffusion processes to the study of fertility rests on three different and somewhat independent developments. In all three cases, the

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

most important contribution is the introduction of conceptual precision and ex ante identification of the mechanisms promoting (inhibiting) diffusion of new behaviors.

An Integrated Theory

In an attempt to grapple with the proper identification of the nature of diffusion processes and adjustment behaviors in fertility, Retherford (1985) formulates an integrated theory that contains many of the advances we singled out as necessary for a testable theory of diffusion. In particular, the author assumes a unique decision-making framework whereby behavioral adjustments (structural factors) and emulation of others’ behaviors (diffusion) may occur in tandem, the latter being more likely in highly integrated communities where psychic costs of deviant behavior are minimized. An important limitation of Retherford’s theory is that it does not contain much elaboration of mechanisms of social influence and only indirect reference to feedback mechanisms.

Coale’s RWA Framework

In a much-cited statement, Coale formulated three preconditions for fertility decline, “ready, willing and able” (RWA). This statement can be the basis for an alternative integrated framework. First, fertility control must be within the field of conscious choice or, equivalently, the new behavior, Bo, must be a member of the set of feasible behaviors among which the individual can choose. A necessary condition for this readiness to exist is that there should be information flowing from members of an individual’s network or from external sources of information. The idea of a new behavior must appear from somewhere. When we refer to ideational change, we seem to have in mind at least this dimension of the process. If so, and as indicated above, ideational change and diffusion should not be used as equivalent concepts because ideational changes may also depend on structural changes. The second and third conditions can be considered simultaneously because they are two parts of a model of rational decision making. Individuals must be willing to engage in the new behavior and they must be able to do so. Being willing refers to the ability to detect net benefits associated with the new behavior—what we referred to earlier as the linkage between net benefits and behavior. Being able simply refers to the accessibility to means to engage in the behavior and to the ability to bypass institutional constraints that impede the practice of the behavior.

Coale’s RWA framework is agnostic regarding the nature of forces that may erode or develop support for each of the three preconditions. In particular, changes in any one of the three components could involve

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

both ideational changes as well as nonideational changes, and all three of them could be affected by diffusion processes to different degrees.

Coale’s integrated framework has been recently operationalized in a number of developing countries by Lesthaeghe and Vanderhoeft (this volume). They assess the status of these three conditions and estimate their influence on the onset and speed of observed fertility changes (Lesthaeghe and Vanderhoeft, this volume). The limitation of this kind of work is that, in order to test diffusion models, one needs to estimate the effects of social influences (and feedbacks) on the level and patterns of each of the three components. Only then could one assess the overall contribution of diffusion to observed fertility changes, and to estimate the relative weight of the influence of the diffusion mechanisms across the conditions contained in the RWA set.

Social Learning, Feedbacks, and Institutional Constraints

Finally, recent developments in model formulation and in empirical analyses have led to important improvements on two fronts. The first consists of defining explicitly an individual-based decision-making process that acknowledges the operation of social influences, and then formulating a model of such a process whose parameters are estimable from available data. Once parameters are estimated, hypothesis testing is carried out to determine if the estimates are what we would expect if social influences were indeed part and parcel of the decision-making process. The bulk of this work has been carried by a few researchers but mainly by Casterline, Montgomery, and Rosero-Bixby in various publications (Rosero-Bixby, 1991; Casterline et al., 1987; Rosero-Bixby and Casterline, 1993; Montgomery and Casterline, 1993, 1996; Montgomery and Chung, 1994). Although this work utilizes different types of models, some more data demanding than others, it derives from a unified framework (see earlier sections) that makes it comparable to other attempts to tease out social influences from observed behaviors either in organizational contexts (Erbring and Young, 1979; Marsden and Friedkin, 1993) or in social movements (Liao, 1994).

One of the shortcomings of these models is that they do not specify the network dynamics in detail, although they allow simplified representation of what social influences are. In a second line of improvements, researchers focus much more rigorously on the actual configuration of networks in which adopters and nonadopters may participate. In particular, the models are formulated to understand the dynamic interplay between individual decision making and the aggregate properties of the system, notably the continuous reshuffling of network connections that take place as the diffusion process advances (Kohler, 1998; Durlauf and Walker, this volume).

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

SUMMARY AND CONCLUSIONS

The main task of this paper has been to derive explicit rules for testing the existence of diffusion processes and their mechanisms or, equivalently, to formulate conditions for the empirical identification of diffusion processes.

I begin by recognizing the opposition between structural and diffusion-based explanations and confirm that this contrast is pervasive in demography and sociology. Furthermore, I also verify that, in most cases, the contrast is ill posed, ill defined, and poorly resolved through empirical analyses. In particular, I suggest that the opposition between the two types of explanations tends to undermine and overlook the decision-making process that is at the root of every diffusion process.

Using previous discussions and elaborations on the subject, I introduce a preliminary, minimal definition that enables identification of key elements of a diffusion process. These are decision making, resistance and thresholds, social influence, rejection, and feedbacks. Armed with this minimal definition, I undertake the task of reviewing broad areas of application of diffusion models in sociology and demography in general, and identify several stages in the history of sociological applications. I discuss recent applications in collective action and organizational theory as examples of what would be near-to-ideal conditions for model formulation and testing of diffusion processes. This review leads me to the elaboration of a much refined definition of diffusion, one that highlights what is unique to a diffusion process, namely, the salience of social influence in decision making, and three mediating mechanisms through which social influence modifies individual behavior. This leads to the formulation of a golden standard or ideal model to uniquely represent and distinguish among various mechanisms of diffusion. Finally, I state fairly precise conditions for empirical identification of such processes.

The paper ends with a brief review of diffusion research in the area of fertility. This review reveals that only very recent applications and hypotheses verification meet the stricter conditions set forth in the previous section’s discussion. Paradigms used in the past are simply too loose, too unspecific, and ultimately too far removed from the golden standard to be considered as anything more than useful suggestions. The most promising areas of research are those that rest on integrated formulations, where changes via diffusion and structural adjustments are viewed as results of individual decision-making processes that include individuals’ social and economic characteristics and individuals’ ties to significant others in a set of relevant social networks. We need refinements in the identification of how individuals choose and remain members of social networks, on the nature of feedback mechanisms, and

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

on the existence of institutional constraints and their effects on individual choices.

The most important conclusion we draw from this review, and one that should guide future efforts in the area of fertility and other social behaviors, is that the sharp divide established between processes of structural adjustment and diffusion may be a good didactic tool and part and parcel of a respectable intellectual tradition, but it is seldom an accurate way of portraying the mechanisms that shape the behaviors they are intended to explain.

NOTES

1.  

An important idea that I will defend later is that one should not conflate the notions of cultural or ideational explanation with the notion of diffusion. They are simply not equivalent, and many confusions could be avoided if we kept them separate.

2.  

An interesting example of a case study of forerunners is Livi-Bacci’s (1986) description of apparent practices of fertility control among elites and other selected social groups in Western Europe.

3.  

Potter (1998) addresses the problem explicitly, though devoting more attention to what he calls “pernicious aspects” of social interactions that end up imparting inertia in the adoption of contraceptive technology and locking populations into a restricted menu of contraceptive choices, and less attention to mechanisms of outright abandonment of an adopted practice. Sinding and Mason (1998) also addresses the problem of rejection and, finally, Kohler’s new work (Kohler, 1997) provides an opportunity for rigorous formal treatment of it. This problem has been better formalized in the literature on collective violence (Myers, 1997).

4.  

See the paper by Durlauf and Walker (this volume) for a formal treatment of some aspects of the endogenous feedback mechanism.

5.  

The selection issues arising from devoting overwhelming attention to diffusion processes that more or less succeed in taking hold, while neglecting those where diffusion never takes off or dies out shortly after its onset, are presumably quite important but, to my knowledge, have not been studied seriously.

6.  

An important aspect of the weaknesses of these models to identify underlying processes is that researchers who employ them usually assess the fit between observed and expected outcomes by examination of cumulative occurrence (proportion of the population who has adopted, for example). It is well known that a good fit of a cumulative distribution can conceal complete failure to predict associated densities (frequencies of new adopters during a small time interval).

7.  

This third mechanism is associated with a number of interesting formal and substantive problems regarding the possibility of unstable equilibria and the relation between small changes at the individual level that may translate into large changes at the aggregate level (see Durlauf and Walker, this volume). As formulated here this mechanism includes what Arthur (1989) identifies as sources of increasing returns that emerge as an adoption process gets under way. Increasing returns can occur due to coordination externalities, advantages associated with learning, and advantages associated with increased information flows. These are all sources of positive feedback. The formulation I suggest here, however, leaves the door open for the possibility that feedback also can be negative.

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

8.  

See examples of the use of spatial proximity in the previous discussion of applications to collective action and organizations.

9.  

Needless to say, controlled experiments, though close to this ideal, are not close enough.

10.  

For a review of processes of network formation, see Doreian and Stokman (1997).

REFERENCES

Aldrich, H.E. 1979 Organizations and Environments. Englewood Cliffs, NJ: Prentice-Hall.

Anderson, R.M., and R.M.May 1992 Infectious Diseases of Humans. Oxford: Oxford University Press.

Anselin, L. 1988 Spatial Econometrics: Methods and Models. Boston: Kluwer.

Arthur, B.W. 1989 Competing technologies, increasing returns, and lock-in by historical events. Economics Journal 99:116–131.


Bailey, N.T.J. 1975 The Mathematical Theory of Infectious Diseases and Its Applications. London: Charles Griffen.

Bandura, A. 1986 Social Foundations of Thought and Action. Englewood Cliffs, NJ: Prentice-Hall.

Bartholomew, D.J. 1982 Stochastic Models for Social Processes. New York: Wiley.

Bocquet-Appel, J.-P. 1997 Diffusion Spatiale de la Contraception en Grande-Bretagne, a l’Origine de la Transition. July, Institut National d’Etudes Demographiques. Seminaire Demodynamiques.

Bongaarts, J., and S.C.Watkins 1996 Social interactions and contemporary fertility transitions. Population and Development Review 22(4):639–682.

Brown, L. 1981 Innovation Diffusion: A New Perspective. London: Methuen.

Burt, R.S. 1987 Social contagion and innovation: Cohesion versus structural equivalence. American Journal of Sociology 92:1287–1335.


Caldwell, J. 1982 Theory of Fertility Decline. New York: Academic Press.

Carlsson, G. 1966 The decline of fertility: Innovation or adjustment process? Population Studies 20: 149–174.

Casterline, J.B., M.R.Montgomery, and R.L.Clark 1987 Diffusion Models of Fertility Control: Are There New Insights? PSTC WP 87–06. July, Brown University.

Cavalli-Sforza, L.L., and M.W.Feldman 1981 Cultural Transmission and Evolution: A Quantitative Approach. Princeton: Princeton University Press.

Cleland, J., and C.Wilson 1987 Demand theories of the fertility transition: An iconoclastic view. Population Studies 41:5–30.

Coale, A.J. 1973 The Demographic Transition. Liège, Belgium: Ordina.

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

Coale, A.J., and S.C.Watkins 1986 The Decline of Fertility in Europe. Princeton: Princeton University Press.

Cohen, B., and M.Montgomery 1998 Introduction. Pp. 1–38 in From Death to Birth: Mortality Decline and Reproductive Change, M.Montgomery and B.Cohen, eds. Committee on Population, Commission on Behavioral and Social Sciences and Education. Washington, DC: National Academy Press.

Coleman, J.S. 1990 Foundations of Social Theory. Cambridge, MA: Harvard University Press.

Coleman, J.S., E.Katz, and H.Menzel 1966 Medical Innovation: A Diffusion Study. New York: Bobbs Merrill.

Cressie, N. 1991 Statistics for Spatial Data. New York: Wiley.

Davis, K. 1963 The theory of change and response in modern demographic history. Population Index 29(4):345–366.

DiMaggio, P.J., and W.W.Powell 1991 The iron cage revisited: Institutional isomorphism and collective rationality in organization fields. In The New Institutionalism in Organizational Quantitative Analysis, P.J.DiMaggio and W.W.Powell, eds. Chicago: University of Chicago Press.

Doreian, P., and F.N.Stokman, eds. 1997 Evolution of Social Networks. Amsterdam: Gordon and Breach.


Erbring, L., and A.A.Young 1979 Individuals and social structure: Contextual effects as endogenous feedback. Sociological Methods and Research 7(4):396–30.


Fligstein, N. 1985 The spread of the multidivisional form among large firms, 1919–1979. American Sociological Review 50(June):377–391.


Galloway, P., E.A.Hammel, and R.D.Lee 1994 Fertility decline in Prussia, 1875–1910: A pooled cross-section time series analysis. Population Studies 48(1):135–148.

Granovetter, M. 1973 The strength of weak ties. American Journal of Sociology 78:1360–1380.

1978 Threshold models of collective behavior. American Journal of Sociology 83:1420– 1443.


Hagerstrand, T. 1967 Innovation Diffusion as a Spatial Process. Chicago: University of Chicago Press.

Hamblin, R.L., R.B.Jacobsen, and J.L.L.Miller 1973 A Mathematical Theory of Social Change. New York: Wiley.

Hannan, M.T., and J.H.Freeman 1977 The population ecology of organizations. American Journal of Sociology 82:929–964.

Hechter, M. 1975 Internal Colonialism: The Celtic Fringe in British National Development, 1536–1966. London: Routledge and Kegan Paul.

Hedström, P. 1994 Contagious collectivities: On the spatial diffusion of Swedish trade unions 1890– 1940. American Journal of Sociology 99(5): 1157–1179.


Jenkins, C.J., and C.M.Eckert 1986 Channeling black insurgency: Elite patronage and professional social movement organizations in the development of black movement. American Sociological Review 51:812–829.

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

Katz, E. 1961 The social itinerary of technical change: Two studies on diffusion of innovation. In Studies on Innovation and of Communication to the Public, W.Schramm, ed. Stanford: Stanford University.

Katz, E., and P.P.Lazarsfeld 1955 Personal Influence: The Part Played by People in the Flow of Mass Communications. New York: Free Press.

Knodel, J. 1974 The Decline of Fertility in Germany. Princeton: Princeton University Press.

Knodel, J., and E.van de Walle 1967 Breast feeding, fertility, and infant mortality. Population Studies 21(2):109–131.

Kohler, H. 1997 Learning in Social Networks and Contraceptive Choice. Demography 34(3):369– 383.

Kohler, H.-P. 1998 Social Interactions and Fluctuations in Birth Rates. Unpublished manuscript. University of California at Berkeley.


Laumann, E.O., P.V.Marsden, and J.Galaskiewicz 1977 Community-elite influence structures: Extension of a network approach. American Journal of Sociology 83:594–631.

LeBon, G. 1897 The Crowd. London: Unwin.

Lesthaeghe, R.J. 1977 The Decline of Belgian Fertility, 1800–1970. Princeton: Princeton University Press.

Liao, T.F. 1994 A theoretical framework of collective action for the evaluation of family planning programs. Population Research and Policy Review 13:49–67.

Livi-Bacci, M. 1971 A Century of Portuguese Fertility. Princeton: Princeton University Press. 1986 Social-group forerunners of fertility control in Europe. In The Decline of Fertility in Europe, A.Coale and S.Watkins, eds. Princeton: Princeton University Press.


Mahajan, V., and R.A.Peterson 1985 Models for Innovation Diffusion. Newbury Park, CA: Sage Publications.

March, J.G., and J.P.Olsen 1976 Ambiguity and Choice in Organizations. Bergen, Norway: Universitetsforlaget.

Marsden, P. 1998 Diffusion Through Social Networks. Unpublished paper presented at Workshop on Social Processes Underlying Fertility Change in Developing Countries, National Research Council, January 29–30, 1998, Washington, DC.

Marsden, P., and N.Friedkin 1993 Network studies of social influence. Sociological Methods and Research 22(1):127– 151.

Marwell, G., and P.E.Oliver 1993 The Critical Mass in Collective Action: A Micro-Social Theory. Cambridge, Eng.: Cambridge University Press.

McAdam, D. 1982 Political Process and the Development of Black Insurgency 1930–1970. Chicago: University of Chicago Press.

1995 Initiator and spin-off movements: Diffusion processes in protest cycles. Pp. 217– 239 in Repertoires and Cycles of Collective Action, M.Traugott, ed. Durham, NC: Duke University Press.

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

McAdam, D., and D.Rucht 1993 The cross national diffusion of movement ideas. American Academy of Political and Social Science 528:56–74, July.

Montgomery, M.R., and J.B.Casterline 1993 The diffusion of fertility control in Taiwan: Evidence from pooled cross-section, time-series models. Population Studies 47(3):457–479.

1996 Social Learning, Social Influence, and New Models of Fertility. New York: The Population Council.

Montgomery, M.R., and W.S.Chung 1994 Social networks and the diffusion of fertility control in Korea. In Cultural and Temporal Variations in Values: Impact on Fertility Change, Richard Leete, ed. Oxford, Eng.: Oxford University Press.

Morris, M. 1993 Epidemiology and social networks: Modeling structured diffusion. Sociological Methods and Research 22:99–126.

Moscovici, S. 1985 Social influence and conformity. The Handbook of Social Psychology 2:347–412.

Myers, D.J. 1997 Diffusion Models for Riots and Other Collective Violence. Unpublished Ph.D. dissertation, University of Wisconsin.

Notestein, F.W. 1945 Population—the long view. Pp. 36–57 in Food for the World, T.W.Schultz, ed. Chicago: University of Chicago Press.


Obershall, A. 1989 The 1960s sit-ins: Protest diffusion and movement takeoff. Research in Social Movements, Conflict and Change 11:31–33.

Okun, B.S. 1994 Evaluating methods for detecting fertility control: Method and cohort parity analysis . Population Studies 48:193–222.

Olson, M. 1965 The Logic of Collective Action. Cambridge, MA: Harvard University Press.

Olzak, S. 1992 The Dynamics of Ethnic Competition and Conflict. Palo Alto: Stanford University Press.


Palloni, A., and H.Rafalimanana 1997 The Effects of Infant Mortality on Fertility Revisited: Some New Evidence. Center for Demography and Ecology Working Paper No. 96–27. University of Wisconsin-Madison.

Pitcher, B.L., R.L.Hamblin, and J.L.L.Miller 1978 Diffusion of collective violence. American Sociological Review 43:23–35.

Piven, F.F., and R.Cloward 1979 Poor People’s Movements: Why They Succeed, How They Fail. New York: Vintage.

Potter, J., R.M.Assumcão, S.M.Cavenghi, and A.J.Caetano 1998 The Spread of Television and Fertility Decline in Brazil: A spatial-temporal analysis: 1970–1991. Unpublished paper presented at Workshop on Social Processes Underlying Fertility Change in Developing Countries, National Research Council, January 29–30, 1998, Washington, DC.

Preston, S.H. 1978 Introduction. In The Effects of Infant and Child Mortality on Fertility, S.H.Preston, ed. New York: Academic Press.

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×

Retherford, R.D. 1985 A theory of marital fertility. Population Studies 39:249–268.

Rogers, E.M. 1962 The Diffusion of Innovations. 1st ed. New York: Free Press.

1973 The Diffusion of Innovations. 2d ed. New York: Free Press.

1983 The Diffusion of Innovations. New York: Free Press.

1988 The Diffusion of Innovations. 3d ed. New York: Free Press.

1995 The Diffusion of Innovations. 4th ed. New York: Free Press.

Rogers, E.M., and D.L.Kincaid 1981 Communication Networks: A New Paradigm for Research. New York: Free Press.

Rosero-Bixby, L. 1991 Interaction Diffusion and Fertility Transition in Costa Rica. Unpublished Ph.D. dissertation, School of Public Health, University of Michigan.

Rosero-Bixby, L., and J.B.Casterline 1993 Modelling diffusion effects in fertility transition. Population Studies 47(1):147–167.

1994 Interaction diffusion and fertility transition in Costa Rica. Social Forces 73(2):435– 462.

Ryan, B., and N.C.Gross 1943 The diffusion of hybrid seed corn in two Iowa communities. Rural Sociology 8:15– 24.


Sinding, S., and K.O.Mason 1998 Diffusion Theories and Population Policies. Unpublished paper presented at Workshop on Social Processes Underlying Fertility Change in Developing Countries, National Research Council, January 29–30, 1998, Washington, DC.

Strang, D. 1991 Adding social structure to diffusion models: An event history framework. Sociological Methods and Research 19:324–353.

Strang, D., and S.Soule 1997 Diffusion in Organizations and Social Movements: From Hybrid Corn to Poison Pills. Technical Report 9702. Department of Sociology, University of Arizona.

Strang, D., and N.B.Tuma 1993 Spatial and temporal heterogeneity in diffusion. American Journal of Sociology 99:614–639.


Tarrow, S. 1994 Power in Movement. Cambridge, Eng.: Cambridge University Press.

Tilly, C. 1978 From Mobilization to Revolution. New York: McGraw-Hill.


Valente, T.W. 1995 Network Models of the Diffusion of Innovations. Cresskill, NJ: Hampton Press,

van de Walle, F. 1986 Infant mortality and the European demographic transition. Pp. 201–233 in The Decline of Fertility in Europe, A.J.Coale and S.C.Watkins, eds. Princeton: Princeton University Press.

Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 66
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 67
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 68
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 69
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 70
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 71
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 72
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 73
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 74
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 75
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 76
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 77
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 78
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 79
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 80
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 81
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 82
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 83
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 84
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 85
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 86
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 87
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 88
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 89
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 90
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 91
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 92
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 93
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 94
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 95
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 96
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 97
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 98
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 99
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 100
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 101
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 102
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 103
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 104
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 105
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 106
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 107
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 108
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 109
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 110
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 111
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 112
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 113
Suggested Citation:"3 Diffusion in Sociological Analysis." National Research Council. 2001. Diffusion Processes and Fertility Transition: Selected Perspectives. Washington, DC: The National Academies Press. doi: 10.17226/10228.
×
Page 114
Next: 4 Social Interactions and Fertility Transitions »
Diffusion Processes and Fertility Transition: Selected Perspectives Get This Book
×
Buy Paperback | $61.00 Buy Ebook | $48.99
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

This volume is part of an effort to review what is known about the determinants of fertility transition in developing countries and to identify lessons that might lead to policies aimed at lowering fertility. It addresses the roles of diffusion processes, ideational change, social networks, and mass communications in changing behavior and values, especially as related to childbearing. A new body of empirical research is currently emerging from studies of social networks in Asia (Thailand, Taiwan, Korea), Latin America (Costa Rica), and Sub-Saharan Africa (Kenya, Malawi, Ghana). Given the potential significance of social interactions to the design of effective family planning programs in high-fertility settings, efforts to synthesize this emerging body of literature are clearly important.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    Switch between the Original Pages, where you can read the report as it appeared in print, and Text Pages for the web version, where you can highlight and search the text.

    « Back Next »
  6. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  7. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  8. ×

    View our suggested citation for this chapter.

    « Back Next »
  9. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!