טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
Ph.D Thesis
Ph.D StudentKaplan Sigal
SubjectDevelopment and Estimation of a Semi-Compensatory Model
with a Flexible Error Structure: Application to
Residential Choice
DepartmentDepartment of Civil and Environmental Engineering
Supervisors Professor Yoram Shiftan
Professor Shlomo Bekhor
Full Thesis textFull thesis text - English Version


Abstract

Semi-compensatory models show promise in representing two-stage choice processes by integrating choice set formation within discrete choice models. However, they are subject to simplifying assumptions that limit their application. This dissertation focuses of the development and estimation of a novel semi-compensatory model that alleviates the need for simplifying assumptions concerning (i) the number of alternatives, (ii) the representation of choice set formation, and (iii) the error structure.

The proposed semi-compensatory model represents a sequence of choice set formation based on the conjunctive heuristic and utility-based choice. A basic model variation and three extensions are formulated. The basic model assumes an i.i.d. error structure, and hence jointly represents the conjunctive heuristic with multiple ordered-probit models and the utility-based choice with the multinomial logit. The model extensions account for correlations across thresholds, nested correlation patterns and random taste variations. Hence, the model extensions represent the conjunctive heuristic with a multidimensional mixed ordered-response probit model and the utility-based choice with alternatively (i) a multinomial logit model, (ii) a nested logit model and (ii) a random coefficients logit model.

The proposed semi-compensatory model is applied to off-campus rental apartment choice of students. A web-based survey including a choice experiment and a questionnaire was designed for data collection. The experiment recorded choice protocols of respondents who searched an apartment inventory by selecting predefined criteria thresholds and chose their three most preferred apartments from the resulting choice set. The questionnaire collected personal information that can be associated to the selected thresholds and choice outcomes.

The data for model estimation consists of 1,893 observations of criteria thresholds and relative choice outcomes from a population sample of 631 respondents. The universal realm contains 200 alternatives. Results show (i) the efficiency of the proposed web-based data collection method in retrieving semi-compensatory choice protocols, (ii) the possibility to apply the proposed model to choice contexts with a large universal realm of alternatives, (iii) the determinants of threshold selection, (iv) the importance of accounting for correlation patterns and random taste variations in semi-compensatory models, and (v) the prediction potential of the model versus the compensatory approach.