Publication Detail
Binary and Multinomial Logit Models of the Preference for Center-Based Telecommuting
UCD-ITS-RR-95-16 Research Report |
Suggested Citation:
Bagley, Michael N. (1995) Binary and Multinomial Logit Models of the Preference for Center-Based Telecommuting. Institute of Transportation Studies, University of California, Davis, Research Report UCD-ITS-RR-95-16
This study examines the individual's preference to telecommute from a center. Empirical data obtained from 628 employees of the City of San Diego were used in the development of three preference models based on a previously developed conceptual model of the decision to telecommute. Two binary logit models are presented, one on the preference to telecommute from a center versus not telecommuting from a center (ρ2 = 0.28), and the other on the preference to telecommute from a center over telecommuting from home (ρ2 = 0.68). A multinomial logit model is also estimated on the following four alternatives: preferring to telecommute from home, preferring to telecommute from a center, preferring not to telecommute, and preferring either form of telecommuting (ρ2 = 0.34). Tests for the independence from irrelevant alternatives (IIA) property on the third model showed that a multinomial logit structure was invalid, and that future research into other model structures such as multinomial probit or nested log should be conducted.
The results of the models unquestionably displayed the importance of attitudinal measures in measuring an individual's preference to telecommute. Indeed, oblique factor scores representing workplace interaction, stress, workaholism, internal control, and commute stress were statistically significant in some or all of the models. Other explanatory variables which were found to be consistently significant were education, job suitability, and age.
The results of the models unquestionably displayed the importance of attitudinal measures in measuring an individual's preference to telecommute. Indeed, oblique factor scores representing workplace interaction, stress, workaholism, internal control, and commute stress were statistically significant in some or all of the models. Other explanatory variables which were found to be consistently significant were education, job suitability, and age.
Master's Thesis.