More linear than you would think : quadratic utilities, product differentiation and trade patterns

The increasing availability of detailed trade datasets is exposing limits in existing trade theories that need to be addressed. For example, this Thesis documents how the observed micro and aggregate trade patterns cannot be explained by the combinations of functional forms and sources of heterogeneity proposed in the existing literature. To fill this gap, it is here developed a new class of space-augmented trade models based on quadratic utilities with asymmetric varieties and heterogeneous consumers yielding rich patterns of multi-dimensional heterogeneity across firms and markets. Different models are analysed, each one associated with specific properties in terms of market outcomes and underlying spatial structure. Hence, based on the observed or desired properties of the problem at hand, trade practitioners can refer to this Thesis to find guidance on which model suits better their research purposes. One model in particular, called the “verti-zontal”, is identified as the most appropriate to study trade patterns in general, when all products are included in the sample. Its validity is tested on micro and aggregate trade data, showing that market and exporter characteristics are enough to capture price variability across markets but an additional, variety-market-specific source of variability is needed to explain quantity variability. The models can be used for research purposes, to identify demand parameters, or for policy purposes, to produce indices of competitiveness at the country-product level.(ECGE - Sciences économiques et de gestion) -- UCL, 201