This thesis compares different approaches to estimating budgets for Kuhn-Tucker (KT) demand systems, more specifically for the multiple discrete-continuous extreme value (MDCEV) model. The approaches tested include: (1) The log-linear regression approach (2) The stochastic frontier regression approach, and (3) arbitrarily assumed budgets that are not necessarily modeled as a function of decision maker characteristics and choice-environment characteristics. The log-linear regression approach has been used in the literature to model the observed total expenditure as way of estimating budgets for the MDCEV models. This approach allows the total expenditure to depend on the characteristics of the choice-maker and the choice environment. However, this approach does not offer an easy way to allow the total expenditure to change due to changes in choice alternative-specific attributes, but only allows a reallocation of the observed total expenditure among the different choice alternatives. To address this issue, we propose the stochastic frontier regression approach. The approach is useful when the underlying budgets driving a choice situation are unobserved, but only the expenditures on the choice alternatives of interest are observed. The approach is based on the notion that consumers operate under latent budgets that can be conceived (and modeled using stochastic frontier regression) as the maximum possible expenditure they are willing to incur. To compare the efficacy of the above-mentioned approaches, we performed two empirical assessments: (1) The analysis of out-of-home activity participation and time-use (with a budget on the total time available for out-of-home activities) for a sample of non-working adults in Florida, and (2) The analysis of household vehicle type/vintage holdings and usage (with a budget on the total annual mileage) for a sample of households in Florida. A comparison of the MDCEV model predictions (based on budgets from the above mentioned approaches) demonstrates that the log-linear regression approach and the stochastic frontier approach performed better than arbitrarily assumed budgets approaches. This is because both approaches consider heterogeneity in budgets due to socio-demographics and other explanatory factors rather than arbitrarily imposing uniform budgets on all consumers. Between the log-linear regression and the stochastic frontier regression approaches, the log-linear regression approach resulted in better predictions (vis-à-vis the observed distributions of the discrete-continuous choices) from the MDCEV model. However, policy simulations suggest that the stochastic frontier approach allows the total expenditures to either increase or decrease as a result of changes in alternative-specific attributes. While the log-linear regression approach allows the total expenditures to change as a result of changes in relevant socio-demographic and choice-environment characteristics, it does not allow the total expenditures to change as a result of changes in alternative-specific attributes.

Spatial transferability of travel demand models has been an issue of considerable interest, particularly for small and medium sized planning areas that often do not have the resources and staff time to collect large scale travel survey data and estimate model components native to the region. With the advent of more sophisticated microsimulation-based activity-travel demand models, the interest in spatial transferability has surged in the recent past as smaller metropolitan planning organizations seek to take advantage of emerging modeling methods within the limited resources they can marshal. Traditional approaches to identifying geographical contexts that may borrow and transfer models between one another involve the exogenous a priori identification of a set of variables that are used to characterize the similarity between geographic regions. However, this ad hoc procedure presents considerable challenges as it is difficult to identify the most appropriate criteria a priori. To address this issue, this thesis proposes a latent segmentation approach whereby the most appropriate criteria for identifying areas with similar profiles are determined endogenously within the model estimation phase, customized for every model type. The end products are a set of optimal similarity measures that link regions to one another as well as a fully transferred model, segmented to account for heterogeneity in the population. The methodology is demonstrated and its efficacy established through a case study that utilizes the National Household Travel Survey (NHTS) dataset for information on weekday activities unemployed individuals within 9 regions in the states of California and Florida engage in. A multiple discrete continuous extreme value (MDCEV) model is developed that simulates the discrete nature of activity selection as well as the continuous nature of activity participation. The estimated model is then applied onto the Austin–San Marcos MSA, a context withheld from the original estimation in order to assess its performance. The performance of the segmented model was then examined vis-à-vis that of other models that are similar to the local region in only one dimension. It is found that the methodology offers a robust mechanism for identifying latent segments and establishing criteria for transferring models between areas. / text