Advances in data analysis and survey design of Discrete Choice Experiments: an application to the evaluation of the military presence in Sardinia

This thesis contributes to the literature on economic valuation with stated preference tools based on Hick’s compensated demand function. Among these methods, Discrete Choice Experiments (DCE) are extensively used for investigating individual preferences and assessing the value of hypothetical projects. Academic efforts over the past few decades have been devoted to improving the ability of choice models to describe behaviour, but some research areas call for further consideration. The present work seeks to enhance the comprehension of preference and process heterogeneity and aims to delve into the design of DCEs. These topics were explored through three case studies focusing on preferences towards the military presence in Sardinia, Italy. Every decision regarding the future of military facilities entails a change in the mixture of benefits and costs experienced by hosting communities and hence needs to be carefully considered given the relevance of this topic in the regional public debate. Chapter 2 investigates preferences for the downsizing of Poligono Interforze del Salto di Quirra. Modelling the socio-psychological roots underlying acceptance of hazardous facilities could better inform assessment of welfare changes, yet this approach is seldom used in project appraisal. This chapter aims at filling this gap, integrating a rich socio-psychological framework with the choice model, resulting in a hybrid choice model with a complex multi-layer structure. Results show that place attachment, sense of community and trust in institutions influence the perception of environmental and health risks, and the perception of economic benefits. The last two factors are negatively correlated and directly influence the utility of the attributes characterising the alternatives, driving the preference toward the proposed scenarios. These findings have important practical implications, giving guidance to calibrate welfare compensations and define policy measures aimed at socio-economic regeneration. Chapter 3 explores the role of alternative decision-making strategies in guiding preferences for the downsizing of Poligono di Capo Teulada and Poligono di Capo Frasca. A finite mixture model is used to allocate respondents to different classes based on decision-making procedures, combining the traditional compensatory behaviour with selection-by-aspects strategies and saliency-led choices. This chapter proposes a latent class model with random parameters (LC RPL) to jointly account for preference and process heterogeneity. Latent class analysis indicates that integrating multiple heuristics in the modelling describes satisfactorily individual behaviour. Findings confirm a superior performance of models integrating process heterogeneity, compared to both the MNL and the RPL models. The LC RPL structure used in this chapter produces more conservative model estimates and favours a richer comprehension of citizen preferences. Chapter 4 aims to test whether and to what extent valence-based attribute framing influences the choice estimates with an application on the regeneration of a former military site (Caserma Ederle, Cagliari). Attribute framing has been investigated only to a limited degree in environmental DCE studies. This chapter contributes to its understanding by focusing on alternative framings for the monetary attribute, using a within-subjects experimental treatment. The reliability of the latter is crucial to calculate welfare change measures. RPL model results show higher disutility for the negatively framed monetary attribute but a less dispersed parameter distribution under the positive treatment. In light of these findings, researchers should acknowledge the relevance of attribute framing, proceeding cautiously when defining attributes and levels of the DCE. Pre-testing different wordings for the payment vehicle is recommended, especially when its choice is not obvious or cannot be based on previous studies.