Use this URL to cite or link to this record in EThOS:
Title: Dynamic discrete choice models : forecasting of competitive events through optimal linear filtering of choice persistence effects
Author: Rajkovic, Ivan
Awarding Body: University of Southampton
Current Institution: University of Southampton
Date of Award: 2020
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Thesis embargoed until 20 Apr 2023
Access from Institution:
This thesis presents a new modelling framework for dynamic discrete decision-making problem settings, in which persistence in preferences, derived from previously made ranked choices, is taken into account. The endorsed framework leverages trends of the revealed preferences to model the evolution of the temporal persistence of unobserved attributes of alternatives, and it effectively incorporates changing choice sets and irregular time durations between the repeated availability of alternatives in consecutive decision events. The new model structure eliminates effect-confounding problems inherent in incumbent models, and it highlights the effects of time duration bias and the unreliability of lower ranked choices on the probabilities of future choice selections. Following a 'post-positivistic' research paradigm, empirical validation of the models in a naturalistic market environment (UK horse-betting markets), which integrates behavioural (decisionmaker-related) and economic (betting-market-related) information sets, is carried out. The proposed methodology centres around a two-stage model structure, which includes elements of the classical Conditional Logit approach, revealed order of preferences, and the Kalman filtering of the latent states, aimed at providing forecasts of choice probabilities. These probabilities are subsequently used for implementation of a Kelly betting strategy, which, together with standard statistical tests of significance, assesses the merits of the modelling approach. In particular, it is shown that a novel Kalman filter algorithm, developed for filter divergence mitigation, outperforms traditional Kalman filtering algorithms. The empirical results and the associated analysis confirm that forecasted trend variables add statistically significant information over public market information (betting odds) and that incorporating trend variable forecasts in a betting strategy yields above-average monetary gains. Analysis of the evidence collected in the study leads to the conclusion that persistence in preference effects are significant and have to be controlled for, in order to mitigate the effects of the considered biases. In a wider context, obtained evidence confirms the propensity of vested decision makers to time duration bias in a revealed preference setup and that importance weighting of the ranked choice data may be used to mitigate the effects of lower ranked alternatives’ unreliability.
Supervisor: Sung, Ming-Chien Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available