Concern for Intenet Users’ Information Privacy Concerns
Despite its tremendous potential, the share of electronic (e- )commerce as a percentage of total commerce
remains small, less than 5 percent as of2009.lnformalion privacy concerns related to doing business on the
Internet have been identified as a major factor hindering the growth of e-comrnerce. Therefore, the author
and his colleagues developed and published a scale to measure Internet users’ information privacy concerns (IUIPC). Based on social contract theory, IUIPe was conceptualized to have three dimensions: collection, control, and awareness. Based on exploratory factor analysis, collection was measured by four items or variables. and control and awareness were each represented by three measured variables. Then. another empirical study was designed. and SEM was used to evaluate the properties of the scale (see Chapter 9). First, a measurement model was estimated using CFA. This model was used to establish the composite reliability. and convergent and discriminant validity of the scale. Then. a structural model was estimated and the nomological validity of the scale was established by demonstrating that the theoretical relationships of IUIPC with constructs such as trust. risk. and behavioral intention were supported by the data
In many instances, marketing researchers must answer a set of interrelated questions. For example. a firm providing services may be interested in the following questions: What variables determine service quality? How does service quality influence service attitude and service satisfaction? How does satisfaction with the service result in patronage intention? How does attitude toward the service combine with other variables to affect intention to patronize the service? Such interrelated questions cannot be examined in a unified analysis by any single statistical technique we have discussed;o far in To answer such questions in a unified and integrated manner, the researcher must make use of structural equation modeling (SEM). SEM can help us assess the measurement properties and test the proposed theoretical relationships by using a single techniqucal For example. based on research, we could postulate that service quality has five dimensions or factors such as tangibility, assurance empathy. Service quality could be depicted as a latent construct that is not directly observed or measured. Rather, service quality is represented by the five dimensions that are observed or measured. SEM can determine the contribution of each dimension in representing service quality, and evaluate how well a set of observed variables measuring these dimensions represents service quality, i.e., how reliable is the construct. We can then incorporate this information into the estimation of the relationships between service quality and other constructs. Service quality has a
direct and positive influence on both attitude and satisfaction toward the service. Service attitude and satisfaction, in turn, determine intention to patronize the service. Thus, service attitude and service satisfaction are both dependent and independent variables in our theory. A hypothesized dependent variable (service attitude/satisfaction) can become an independent variable in a subsequent dependence relationship (explaining patronage intention). Later in the chapter. we give an empirical application of service quality in the context of banking.
SEM examines the structure of these interrelationships, which are expressed in a series of structural equations. This concept is similar to estimating a series of multiple regression equations These equations model all the relationships among constructs, dependent as well as independent. In SEM, the constructs are unobservable or latent factors that are represented by multiple variables. This is similar to the concept of variables representing a factor in factor analysis but SEM explicitly takes into account the measurement error. Measurement error is the degree to which the observed variables do not describe the latent constructs of interest in SEM. SEM is distinguished from other multivariate techniques we
have discussed in by the following characteristics.”
1. Representation of constructs as un observable or latent factors in dependence relationships.
2. Estimation of multiple and interrelated dependence relationships incorporated in an integrated model.
3. Incorporation of measurement error in an explicit manner. SEM can explicitly account for less than perfect reliability of the observed variables. providing analyses of attenuation and estimation bias due to measurement error.
4. Explanation of the covariance among the observed variables. SEM seeks to represent hypotheses about the means, variances, and covariances of observed data in terms of a smaller number of structural parameters defined by a hypothesized underlying model.
Foundations of SEM
Foundational to the understanding of SEM are the concepts of theory, model, path diagram, exogenous versus endogenous constructs, dependence and correlation relationships, model fit, and model identification. These fundamental concepts are discussed next.
Theory Model and Path Diagram
1.The role of theory and models in developing an approach to the problem was discussed in Chapter
2. There we defined a theory as a conceptual scheme based on foundational statements, or axioms, that are assumed to be true. A theory serves as a conceptual foundation for developing a model. It is very important that an SEM model be based on a theory because all relationships must be specified before the SEM model can be estimated. In SEM, models are often constructed to test certain hypotheses derived from the theory. An SEM model consists of two models: the measurement model and the structural model,” The measurement model depicts how the observed (measured) variables represent constructs. It represents the theory that specifies the observed variables for each construct and permits the assessment of construct validity The observed variables are measured by the researcher and are also referred to as measured variables, manifest variables, indicators, or items of the construct. Typically, observed variables are assumed to be dependent upon constructs’? Thus, straight arrows are drawn from a construct to the observed variables that are indicators of the construct (Figure 22.1). No single indicator can completely represent a construct but is used as an indication of that construct. The measurement model uses the technique of analysis (CFA) in which the researcher specifies which variables define each factor).
It seeks to confirm if the number of factors r constructs) and the loadings of observed (indicator) variables on them conform :0 what is expected on the basis of theory. Thus, CFA is used to verify the factor structure of a set of observed variables. CFA allows the researcher 10 test the hypothesis
that a relationship between observed variables and their underlying latent constructs exists. The researcher uses knowledge of the theory, empirical research, or both; postulates the relationship pattern a and then tests the hypothesis statistically. Indicator variables are selected on the basis of theory, and CFA is used to see if they load as on the expected number of factors. The terms construct and factor are used interchangeably. In other words, in testing for the measurement model, the researcher has complete control over which indicators describe each construct. On
the other hand, a structural model shows how the constructs are interrelated to each other, often with multiple dependence relationships. It specifies whether a relationship exists or does not exist. If a relationship is hypothesized by the theory, then an arrow is drawn. If a relationship is not hypothesized, then no arrow is drawn.
A model is portrayed in a graphical form known as a path diagram. The following norms are followed in constructing a path diagram for a measurement model. Constructs are represented by ovals or circles while measured variables are represented by squares. Straight arrows are drawn from the constructs to the measured variables, as in Figure 22.1(a). Dependence. relationships are portrayed by straight arrows (Figure 22.1 (a» and correlational relationships by curved arrows (Figure 22.1(b»
The process of conducting SEM is described in Figure 22.2. The steps involved in conducting SEM are (I) define the individual constructs. (2) specify the measurement model, (3) assess measurement model reliability and validity, (4) specify the structural model if the measurement model is valid, (5) assess structural model validity, and (6) draw conclusions and make recommendations if the structural model is valid. We will describe each of these steps and discuss the relevant issues involved.
Define the Individual Constructs
As already mentioned, it is very important that SEM analysis be grounded in theory. The specific constructs, how each con truer will be defined and measured, and the interrelationships among constructs must all be specified based on theory. Generally, the interest in SEM is to test both the measurement theory and the structural thcory. Measurement theory specifies how the constructs are represented; structural theory posits how the constructs are interrelated. Structural relationships posited by theory are converted to hypotheses that are then tested using SEM. The test of these hypotheses will be valid only if the underlying measurement model specifying how these constructs are represented is valid. Hence, great care should be taken in ope rationalizing, measuring, and scaling the relevant variables as identified and defined by theory. The measurement and scaling considerations involved, including the development of multi-items scales, were discussed in This process results in scales used to measure the observed variables or indicators