Modeling the Preference for Telecommuting: Measuring Attitudes and Other Variables

This paper begins to operationalize a previously published conceptual model of the individual decision to telecommute. Using survey data from 628 employees of the City of San Diego, hypothesized drives to telecommute and constraints on / facilitators of telecommuting are measured. A binary logit model of the preference to telecommute from home is estimated, having a p 2 of 0.68. The explanatory variables include attitudinal and factual information. Factor analysis is performed on two groups of attitudinal questions, identifying a total of 17 (oblique) factors which can be classified as drives and constraints. Additional measures are created from other data in the survey, usually objective sociodemographic characteristics. Variables representing at least four of the five hypothesized drives (work, family, independence/leisure, and travel) are significant in the final model. Variables from four of the 10 groups of constraints (job suitability, social/professional and household interaction concerns, and a perceived benefit of commuting) are significant, primarily representing internal rather than external constraints. The results clearly demonstrate the importance of attitudinal measures over sociodemographic ones, as the same demographic characteristics (such as the presence of children, commute time) will have different effects on preference for different people. Modeling the Preference for Telecommuting: Measuring Attitudes and Other Variables Patricia L. Mokhtarian and Ilan Salomon INTRODUCTION Telecommuting is often suggested to be one of a series of public policy measures useful in reducing driving and its externalities. In an information age, when significant shares of the workforce are engaged in information processing, the use of telematics as a substitute for travel seems to be a desired and increasingly feasible alternative. Telecommuting is officially included in the list of optional measures to increase (effective) vehicle occupancy ratios; in California and is on the agenda in many other places around the world. As attractive as telecommuting sounds, it seems that forecasts of its adoption (e.g. Nilles, 1988; Boghani et al., 1991) have been over-optimistic (Gold, 1991). Nilles (1988), example, has forecasted that by 1995, under the nominal growth rate, there will be 21.2 millkm telecommuters in the United States, 16% of the workforce or 28% of all information workers. This gap between expectations and reality has warranted deeper analysis as to why people refrain from telecommuting, even when seemingly offered the opportunity to do so. This paper is part of an effort to model the preference for and choice of telecommufing. Herein, we attempt to understand the formation of the preference for telecommufing, which is a precursor of choice. We do not attempt to forecast future preferences, but we present a model of preference in order to identify the factors which affect it. A conceptual model of telecomrnuting adoption developed by the authors describes the relationships among drives, constraints and facilitators as the elements which form preference and choice of telecornrnuting. The model is briefly discussed in the next section. To empirically test these relationships, data was collected from 628 employees of the City of San Diego, who were queried about their perceptions of telecommuting, their own experience (if any) with telecommuting, and attitudes towards work, travel, technology, family and ideology. In this exploratory stage of research into the telecommuting adoption process, the identification of the variables most relevant to that process, and the most effective ways of measuring those variables, is still uncertain. A focus of the research described here has been the development and testing of a large number of potentially-useful measures of explanatory variables of telecommufing preference and choice. Development of those measures, and of their hypothesized impact on telecommuting adoption, has followed the conceptual model described below. Once the pool of candidate explanatory variables had been created, empirical statistical measures were used to select the variables which collectively were believed to have the greatest power to explain telecommuting preference. The high proportion of irfformation explained by the final model (0.68) is taken as general validation of the underlying conceptual constructs and the variable measurement methodology, although refinements in both are clearly possible~ The organization of this report is as follows: Section 2 presents a brief description of the model, the research design and the data, followed by Section 3 which discusses the factor analyses and classifies the resulting factors as drives or constraints. Section 4 describes additional dr/re and constraint variables created from other elements of the survey. Section 5 presents the results of the estimation of telecommuting preference models. The conclusiom and implications for further research are presented in Section 6. 2. THE RESEARCH DESIGN 2,1 The conceptual model The model of telecommuting behavior (described in detail in Mokhtarian and Salomon, 1994, and depicted in a simplified form in Figure 1) suggests that the internal decisionmaking process in which telecommuting is being considered is initiated by some threshold level of dissatisfaction with one or more aspects of life. Such a search for solutions is motivated by "drives". Drives are constructs which "define" aspects of the long-term objectives a person aspires to accomplish, such as an orientation towards family, work, leisure and travel. If there is some dissatisfaction, it is the drives which activate a search for adjustments to reduce the dissatisfaction. Drives, by definition, are "positive" constructs, in the sense that they point to the desired direction of change. When a search for improvement is initiated, information actively or passively received from the environment